19-task plan covering database migration, Import model,
ImportStore refactor, component updates, Filament history page,
failed rows CSV download, and test updates.

Add columns for entity_type, status, headers, column_mappings, results,
failed_rows_data, and row counters. Make file_path and importer nullable
to support the new import wizard workflow where metadata is stored in the
database instead of the filesystem.

…etadata

Refactor all wizard components, steps, and jobs to use the Import
Eloquent model for metadata (status, headers, column_mappings, results,
team_id, user_id) and the slimmed ImportStore for SQLite-only operations
(query, connection, ensureProcessedColumn).

Key changes per file:
- WithImportStore: provides both import() and store() accessors
- UploadStep: creates Import record first, then ImportStore for SQLite
- ImportWizard: loads Import model for restore/status sync
- MappingStep: reads/saves mappings via Import model
- ReviewStep: uses Import for validation dispatch and mappings hash
- PreviewStep: uses Import for status, results, and transition
- ExecuteImportJob: loads both Import and ImportStore, writes results
  and failed rows to DB, creates FailedImportRow records
- ValidateColumnJob: loads Import for metadata, store for SQLite
- ResolveMatchesJob: passes Import to MatchResolver
- MatchResolver: accepts Import as constructor dependency

… architecture

- Rewrite CleanupImportsCommand to query Import records from DB instead of reading meta.json files
- Update all Job tests (ExecuteImportJob, ResolveMatchesJob, ValidateColumnJob) to use Import model
- Fix UploadStepTest to use Import model for metadata and ImportStore for SQLite queries
- Update helper functions to create Import records first, then ImportStore
- Remove ImportStore::load teamId parameter across all tests
- Replace store metadata calls with Import model property access (status, results, headers, etc.)

Updated 4 Livewire test files to use the new ImportStore API:
- ImportWizardTest.php
- MappingStepTest.php
- PreviewStepTest.php
- ReviewStepTest.php

Key changes:
- Import::create([...]) creates DB record, ImportStore::create($import->id) creates SQLite
- ImportStore::load($id) no longer takes teamId parameter
- Metadata methods (setHeaders, setStatus, etc.) moved from ImportStore to Import model
- MatchResolver now takes ($store, $import, $importer) instead of ($store, $importer)
- Match resolution batch ID stored in cache, not in Import model

All 241 ImportWizard tests now passing.

Add integration test verifying Import record fields (status, completed_at,
created_rows, updated_rows, skipped_rows, results) are correctly persisted
after ExecuteImportJob completes. Apply Pint formatting fixes to
ExecuteImportJob and FailedImportRow.

- writeFailedRowsToDb now stores actual raw_data from SQLite rows instead
  of only row_number, making CSV download useful for diagnosing failures
- Generate ULIDs for failed_import_rows primary keys since insert()
  bypasses HasUlids trait
- Strip raw data from failed_rows_data JSON summary on Import model
- DownloadFailedRowsController uses Import headers for CSV columns
  instead of deriving from first row's data keys
- Signed route URLs now expire after 1 hour via temporarySignedRoute

Three-phase plan covering data model cleanup (drop 6 dead columns,
add failed_rows), job safety (failed handler, incremental error
writes, remove silent 100-error cap), and UX state fixes.

Drop file_path, importer, processed_rows, successful_rows, results,
and failed_rows_data columns that were written but never read. Add the
missing failed_rows integer column. Update ExecuteImportJob, Import
model, PreviewStep, ImportHistory, and all tests to use individual
counter columns instead of the JSON results blob.

…cuteImportJob

Fixes three safety issues: (1) adds failed() handler so imports get marked
Failed when the job exhausts retries, (2) flushes failed rows per chunk
alongside other flush operations instead of only on the success path,
(3) removes MAX_STORED_ERRORS cap that silently discarded errors beyond 100.

Prevent users from navigating back to MappingStep while validation
batches are running, which caused stale data corruption. ReviewStep
now dispatches validation-state-changed events to ImportWizard.

Replace property-based mappings hash tracking with cache-based
validation state. Remove validationInProgress from ImportWizard
since backward navigation during validation is now allowed.
Cancel old batches on re-entry and skip dispatching when cache
hash matches.

Apply rector automated refactoring: query builder over magic
methods, scope attributes, instanceof checks, first-class
callables, abort_if, remove unused params. Fix all 38 phpstan
errors by adding @Property annotations to models, guarding
fopen resource types, and removing dead constructor properties.