…ization (#8087)

When you export shapes as SVG, paste back (creates image shape), then
export again with other shapes — the second export wraps the first SVG
as `<image href="data:image/svg+xml;base64,...">`. On paste,
`sanitizeSvg()` strips this href because `sanitizeUri` only allows
raster data URIs on `<image>` elements. Result: blank image.

This PR decodes the embedded SVG, runs `sanitizeSvg()` recursively
(strips scripts/XSS), then re-encodes. Fully malicious embedded SVGs (no
safe content left) are still blocked. Also adds a `MAX_NODE_DEPTH=100`
cap on `sanitizeNode` recursion to prevent DoS from deeply nested
elements.

### Change type

- [x] `bugfix`

### Test plan

1. Export a geo shape as SVG
2. Paste it back (creates image shape)
3. Select that image shape + another shape, export as SVG
4. Paste the resulting SVG — the image should render correctly, not be
blank

- [x] Unit tests

### Release notes

- Fixed SVG sanitizer stripping embedded SVG data URIs on `<image>`
elements. Nested SVGs (from re-exporting pasted SVG images) are now
recursively sanitized instead of blocked.

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Medium Risk**
> Touches XSS-sensitive SVG sanitization logic and introduces recursive
decoding/encoding of data URIs, so subtle parsing/encoding or missed
edge cases could create security or rendering regressions.
> 
> **Overview**
> Fixes a regression where exported SVGs containing nested `<image
href="data:image/svg+xml;base64,...">` were being blanked by the
sanitizer.
> 
> `sanitizeSvg` now allows `data:image/svg+xml` on `<image>`/`<feImage>`
by decoding the embedded SVG, re-running sanitization recursively, and
re-encoding the sanitized result (dropping the `href` entirely if the
embedded SVG has no safe content). Adds an embed recursion depth limit
and new unit tests covering safe embedded SVG preservation and
script-stripping behavior.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
8297037. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

…ive sanitization (#8088)

This is an automated hotfix PR for dotcom deployment.

**Original PR:** [#8087](#8087)
**Original Title:** fix(tldraw): allow data:image/svg+xml on <image>
with recursive sanitization
**Original Author:** @MitjaBezensek

This PR cherry-picks the changes from the original PR to the hotfixes
branch for immediate dotcom deployment.

/cc @MitjaBezensek

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Medium Risk**
> Touches XSS-hardening logic and changes which `data:` URIs are
permitted on `<image>`, so mistakes could reintroduce injection vectors
or break image rendering; recursion depth limiting reduces worst-case
risk.
> 
> **Overview**
> `sanitizeSvg` now allows `data:image/svg+xml` URLs on
`<image>`/`<feImage>` by **decoding and recursively sanitizing** the
embedded SVG, then re-encoding it as a data URI; unsafe/invalid embedded
SVG (or overly deep nesting) causes the `href` to be removed.
> 
> This introduces helpers for data-URI decode/encode, threads a
recursion `depth` through node/attribute sanitization, and expands tests
to cover preserving safe embedded SVG, stripping scripts from
mixed-content embeds, and continuing to block fully malicious embeds.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
aa36002. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

Co-authored-by: Mitja Bezenšek <[email protected]>