…punch, and utils into Block Editor iframe

…ssful clone

- getEligibleTinyMCEFields: filter repeater template fields and _id_
  placeholder textareas from any initialization pass
- escapeSelectorValue: CSS.escape polyfill for safe data-block selector
  building
- selectBestFormInstance: score candidate forms by connected/visible/
  aria-hidden/repeater-template; DOM-order tie-breaker (-index/1000)
- getTinyMCEFieldsFromForms: delegate multi-form case to
  selectBestFormInstance, then prefer visible forms
- resolvePostMessageForms: narrow sowbBlockFormInit target using the
  is-selected + data-block block before falling back to
  selectBestFormInstance
- sanitizeIdSegment: bracket-notation / special-char sanitizer for
  stable textarea ID generation
- resolveWpEditor: single helper that prefers wp.oldEditor over
  wp.editor; shared by all init and teardown paths
- sowbTinyMCEEventNamespace: shared namespace string for all
  document-level on/off calls

Bug fixes:
- siteEditorAddMediaOverride: null-check window.tinymce and editor
  before insertContent/save/fire to avoid crash when tinymce is absent
- hasHealthyTinyMCEEditor: use getElement/getContainer to verify the
  editor actually belongs to this field, not just that it exists
- removeStaleTinyMCEFieldState: use resolveWpEditor() instead of
  duplicating the oldEditor vs editor resolution inline
- setupTinyMCEField: guard window.top.tinyMCEPreInit before accessing
  .mceInit; wrap ed.save()/trigger('change') in if(ed); re-resolve
  wpEditor fresh in wpautopToggleField handler and visibility poll to
  avoid stale closures; use window.wp.oldEditor directly in the
  wp.editor mutation block

Lifecycle hardening:
- clearTinyMCEFieldPendingSetup: also cancel sowb-pre-init-visibility-poll,
  sowb-tinymce-init-timeout and clear the initializing lock data keys
- setupTinyMCEField: add sowb-tinymce-initializing re-entrancy lock with
  a 5 s safety timeout; generate stable IDs via sanitizeIdSegment +
  collision detection instead of Math.random(); deep-clone editorSettings
  to prevent mutation; namespace wp-before-tinymce-init and
  tinymce-editor-setup listeners per field ID so they don't accumulate
- setupTinyMCEFieldInitializer: add 250 ms pre-init visibility poll that
  clears itself when visible; guard repeater template and _id_ fields
- setupSiteEditorTinyMCEFields: accept optional $targetFields so the
  postMessage handler can scope init to the affected form's fields
- message handler: resolve target form via resolvePostMessageForms /
  getTinyMCEFieldsFromForms; guard with sowbTinyMCEMessageListenerBound
  to prevent duplicate listener registration; namespace on/off calls
- Remove setupTinyMCERenderWatchdog (was a no-op poll)

admin.js:
- triggerVisibleRepeaterVisibilityFieldSetup: collect all matching
  fields (visible + hidden) before filtering; fire sowsetupformfield
  only on visible fields; always trigger sowrepeaterfieldsadded on
  document with all fields + clientId so iframe listeners can bind
  pre-init handlers before fields become visible

widget-block.js:
- initializeFormFieldsInIframe: replace burst of 5 scheduled
  postMessages with a single immediate send + persistent
  sowrepeaterfieldsadded listener; listener is namespaced per
  clientId and filters out events from other blocks via originClientId
  comparison; cleanup function removes the listener on unmount
- iframeFormInitKeyRef fingerprint: sample both ends of widgetFormHtml
  (length + slice(0,32) + slice(-32)) instead of storing the full blob
- sowbGetBlockForm / formSelector: use compound selector
  .siteorigin-widget-form.siteorigin-widget-form-main to exclude the
  React wrapper div which only carries siteorigin-widget-form-main
- sowbSetupWidgetForm: trigger sowsetupform on the resolved form after
  initialization so iframe field scripts are notified
```> admin.js:
- triggerVisibleRepeaterVisibilityFieldSetup: collect all matching
  fields (visible + hidden) before filtering; fire sowsetupformfield
  only on visible fields; always trigger sowrepeaterfieldsadded on
  document with all fields + clientId so iframe listeners can bind
  pre-init handlers before fields become visible

widget-block.js:
- initializeFormFieldsInIframe: replace burst of 5 scheduled
  postMessages with a single immediate send + persistent
  sowrepeaterfieldsadded listener; listener is namespaced per
  clientId and filters out events from other blocks via originClientId
  comparison; cleanup function removes the listener on unmount
- iframeFormInitKeyRef fingerprint: sample both ends of widgetFormHtml
  (length + slice(0,32) + slice(-32)) instead of storing the full blob
- sowbGetBlockForm / formSelector: use compound selector
  .siteorigin-widget-form.siteorigin-widget-form-main to exclude the
  React wrapper div which only carries siteorigin-widget-form-main
- sowbSetupWidgetForm: trigger sowsetupform on the resolved form after
  initialization so iframe field scripts are notified

…omments

Fix several issues and add inline comments throughout
tinymce-field.js and widget-block.js to make the non-obvious behaviour
easier to understand for future readers — human or LLM.

tinymce-field.js:
- resolvePostMessageForms: remove dead `is-selected`/`data-block`/
  `aria-selected` selectors. These attributes belong to the parent
  document's block list and are never present in the iframe DOM where
  this function runs; the fallback selectBestFormInstance path was
  already sufficient.
- setupTinyMCEField: add comments explaining the two
  clearTinyMCEFieldPendingSetup calls — the first tears down a previous
  unhealthy init inside the data-initialized branch; the second
  unconditionally clears any lingering pre-init state.
- setupTinyMCEFieldInitializer: clear data-initialized when the editor
  is unhealthy before falling through to re-init. Previously leaving it
  set caused setupTinyMCEField to attempt teardown of an editor that no
  longer exists.
- sowsetupform iframe listener: replace filter().add(find()) with a
  plain find(). The form wrapper element is never itself a TinyMCE
  field, so the filter() was unnecessary and could produce duplicates.
- Drop the $( function() ) jQuery-ready wrapper around the initial
  setupSiteEditorTinyMCEFields() call inside the frameElement branch —
  the script already runs after the iframe DOM is ready.

widget-block.js:
- initializeFormFieldsInIframe: add a comment that the message object
  is intentionally built once from stable closure values and reused
  across all sendInitMessage() calls.
- iframeFFix several correctness issues and add LLM-oriented inline comments
throughout tinymce-field.js and widget-block.js to make non-obvious
behaviour easier to understand for future readers — human or LLM.

tinymce-field.js:
- resolvePostMessageForms: remove dead `is-selected`/`data-block`/
  `aria-selected` selectors. These attributes belong to the parent
  document's block list and are never present in the iframe DOM where
  this function runs; the fallback selectBestFormInstance path was
  already sufficient.
- resolvePostMessageForms: update JSDoc to match the simplified
  implementation — remove the stale clientId param and explain why
  clientId-based narrowing is not done inside the iframe.
- escapeSelectorValue: remove entirely. Its only call site was the
  is-selected/data-block selector block that was removed above; no
  dynamic CSS selectors are built from clientId anywhere in this file.
- setupTinyMCEField: move `$field.attr('data-initialized', true)` to
  after the resolveWpEditor() null-check. Previously a transient null
  wpEditor caused an early return that left the attr set, forcing the
  next call into an unnecessary unhealthy-editor teardown cycle.
- setupTinyMCEField: add comments explaining the two
  clearTinyMCEFieldPendingSetup calls — the first tears down a previous
  unhealthy init inside the data-initialized branch; the second
  unconditionally clears any lingering pre-init state.
- setupTinyMCEField: namespace the media button click handler with
  `.sowbMedia` so that reinit does not accumulate duplicate listeners on
  the same button element.
- setupTinyMCEField: namespace the $wpautopToggleField change handler
  with fieldEventNamespace. The previous bare .off('change') silently
  removed all change listeners on that checkbox, not just ours.
- setupTinyMCEField: in the 5 s safety timeout, also remove the
  `tinymce-editor-setup` listener from window.top.document. Without
  this, fields whose TinyMCE init event never fires (e.g. removed from
  DOM) permanently leak a listener on the top-level document.
- setupTinyMCEFieldInitializer: clear data-initialized when the editor
  is unhealthy before falling through to re-init. Previously leaving it
  set caused setupTinyMCEField to attempt teardown of an editor that no
  longer exists.
- sowsetupform iframe listener: replace filter().add(find()) with a
  plain find(). The form wrapper element is never itself a TinyMCE
  field, so the filter() was unnecessary and could produce duplicates.
- Drop the $( function() ) jQuery-ready wrapper around the initial
  setupSiteEditorTinyMCEFields() call inside the frameElement branch —
  the script already runs after the iframe DOM is ready.

widget-block.js:
- initializeFormFieldsInIframe: add a comment that the message object
  is intentionally built once from stable closure values and reused
  across all sendInitMessage() calls.
- initializeFormFieldsInIframe: add a comment explaining that $fields
  passed with sowrepeaterfieldsadded is intentionally unused — admin.js
  provides it for other listeners, but here a full-form postMessage is
  sent rather than targeting individual fields.
- iframeFormInitKeyRef effect: add a comment explaining that the ref is
  deliberately not updated when initializeFormFieldsInIframe returns
  null (iframe not yet reachable), so the effect retries on the next
  render cycle rather than skipping it permanently.

The inline comments are written to give sufficient context for LLMs
making future changes to understand why the code is structured the way
it is, reducing the risk of well-intentioned but incorrect simplification.ormInitKeyRef effect: add a comment explaining that the ref is
  deliberately not updated when initializeFormFieldsInIframe returns
  null (iframe not yet reachable), so the effect retries on the next
  render cycle rather than skipping it permanently.

The inline comments are written to give sufficient context for LLMs
making future changes to understand why the code is structured the way
it is, reducing the risk of well-intentioned but incorrect simplification.

The _tinymceInitPending[id] entry was constructed using a shared let _resolveInitPending variable in the enclosing function scope. The Promise constructor assigned to it asynchronously, and the .resolve() method on the entry called the variable rather than the captured value. If two editor instances began initialising concurrently, the second assignment would overwrite _resolveInitPending before the first entry's .resolve() had a chance to close over its own value — causing the first editor's pending promise to be resolved (and deleted) by the second editor's resolver instead.

To fix this, the shared-variable pattern was replaced with an IIFE that creates a dedicated storedResolve binding per entry. Each entry now closes over its own resolver; concurrent initialisations can no longer interfere.

…ug fixes

Widget forms rendered inside the Site Editor's editor-canvas iframe were
not correctly initialized due to several compounding issues.

compat/block-editor/widget-block.js:

- Wrong jQuery instance in sowbSetupWidgetForm. The form node returned by
  sowbGetBlockForm was wrapped in the parent-document jQuery, but
  sowSetupForm(), wpColorPicker, and all field plugins register against the
  iframe's jQuery. Fixed by re-wrapping the form node via sowbGetElementWindow
  and using the resulting formJQuery/formForms instances throughout.

- Wrong form node selected by sowbGetBlockForm. The PHP form() method renders
  its own .siteorigin-widget-form-main[data-class] element inside the React
  outer wrapper (also .siteorigin-widget-form-main). The outer wrapper has no
  direct .siteorigin-widget-field children, so sowSetupForm found 0 fields —
  breaking sections, repeaters, color pickers, and all other field setup.
  Fixed by narrowing the selector to [data-class], with .first() for safety.

- Duplicate change handlers. Multiple React render cycles could call
  sowbSetupWidgetForm for the same block, stacking unbounded .on('change')
  listeners. Fixed by namespacing with clientId and calling .off() first.

- Added sowbGetEditorCanvasFrame() helper to centralise iframe resolution
  (Site Editor canvas, post-editor canvas, window.frameElement).

- Removed window.frameElement guard from the IIFE's sowbMaybeSetupSiteEditorAssets
  call — the function resolves the correct frame internally.

base/inc/fields/js/tinymce-field.js:

- _tinymceInitPending entry stranded on cleanup. clearTinyMCEFieldPendingSetup
  cancelled the 5 s safety timeout — the only fallback resolver — without
  resolving the associated promise, leaving sowbGetTinyMCEInitPromise() waiting
  indefinitely. Fixed by resolving and deleting the pending entry before wiping
  sowb-tinymce-initializing-id.

- data-initialized cleared during an in-flight init. setupTinyMCEFieldInitializer
  checked hasHealthyTinyMCEEditor (returns false while TinyMCE is still
  initializing), cleared data-initialized, then returned immediately because
  sowb-tinymce-initializing was set. When the original init completed the field
  was left unmarked, causing the next setup event to tear down a healthy editor.
  Fixed by returning early when sowb-tinymce-initializing is already set.

- sowbBlockFormInit postMessage handler simplified. The sowSetupForm call
  previously added to the handler is removed; sowbSetupWidgetForm already runs it
  with the correct jQuery instance before the postMessage is ever sent. The handler
  now covers only TinyMCE field initialization.

…acy-handler

Block Editor: fix direct widget preview and save persistence

…rame-asset-clone-no-polling

Feature/sowb block editor iframe asset clone no polling