… indexes

…fore remove event

…n api/index

… tags

- TagInput force-fetches document when tags not populated (cache hit fix)
- TagInput updates document.tags observable immediately after add/remove
- Search page fetches all tags on mount so #tag queries resolve tag IDs

- Made query optional (was required non-empty) to allow pure tag searches
- Added tagIds field to DocumentsSearchTitlesSchema
- Pass tagIds through to SearchHelper.searchTitlesForUser in route handler
- SearchHelper.buildWhere already handled tagIds, no changes needed there

With 'search titles only' on and a pure #tag query (no text),
cleanQuery is empty — consistent with how an empty plain-text
search returns nothing.

- Arrow keys navigate suggestions and update the input with the full tag name
- Mouse hover highlights suggestion and fills the input
- Leaving hover restores the typed text
- Enter selects the keyboard-highlighted item, then the first suggestion,
  then creates a new tag (only if no suggestions match)
- Blur restores the typed text if the dropdown closes without selection

Partial tag tokens like #eng that don't resolve to a tag ID were being
treated as an empty filter, returning all documents. Now returns an
empty result set immediately.

Also disables browser autocomplete on the search input to prevent stale
partial tag tokens from appearing as native suggestions.

When clicking a dropdown suggestion, onBlur fired before handleAddTag's
async operations completed, causing the onBlur timeout (150ms) to
restore typedValue to the input. Prevent default on onMouseDown so the
input retains focus and no blur/restore cycle occurs.

- tags API: cross-team isolation, cascade delete, documentCount
  accuracy, viewer auth on add/remove, cross-team boundary on add
- parseSearchQuery: lone hash, hyphens in names, deduplication
- TagsStore: clear() empties name cache, duplicate add doesn't corrupt cache
- documents.info: response includes tags array

Also fixes two bugs found by the new tests:
- parseSearchQuery regex now matches tag names containing hyphens
- parseSearchQuery deduplicates repeated tag tokens

- Handle UniqueConstraintError in tags.update with proper ValidationError
- Only fire tags.remove event when a row is actually destroyed
- Use instance destroy instead of static destroy in tags.remove
- Clean up documentCount type cast in tags.list
- Apply Prettier formatting to all tags feature files

… correctness, tag resolution)