perf: profile and optimize Browse drag-and-drop for large collections

[devin-ai-integration]

Summary

Closes #981. Part of epic #975. Stacked on #996.

Profiled Browse tile dragging at production scale (production build + reorder fixture, real pointer drags via Playwright/CDP — harness committed as scripts/profile_reorder_drag.py, full methodology and numbers in docs/reorder-performance.md).

Diagnosis: long-task cost scaled with tile count and clustered at drag start/end — handleDragStart/handleDragEnd set activeItem (grid state), re-rendering all mounted tiles (600+ MUI cards) because tiles were re-created inline each grid render. Steady-state pointermove/collision cost was not dominant.

Fix (memoization only — zero collision/drop-zone/activation changes, docs/drag-and-drop.md contract untouched):

• ImageTile, CategoryTile → React.memo
• SortableTileGrid: tiles render through a memoized GridTile with stable useCallback render props, so a tile re-renders only when its own item/index/disabled change

Measured (600-image gallery scope): drag FPS 43 → 50; long tasks 2/853 ms/486 ms → 1/368 ms; slowest pointer event 184 ms → 104 ms; heap 100 → 61 MB. The single remaining long task is dnd-kit's one-time shape measurement at drag activation; steady-state dragging has no long tasks.

Compact reorder mode: not implemented — the card grid meets responsiveness criteria at 600+ items after memoization (decision recorded in docs/reorder-performance.md).

Testing

• New frontend/tests/components/tileMemoization.test.tsx: parent re-renders with stable props do not re-execute ImageTile/CategoryTile; prop changes still re-render.
• Full frontend suite: 89 files / 1371 passed (3 expected fail); production build clean.
• Baseline vs post-change profiling runs documented in docs/reorder-performance.md.
• Pre-existing lint error in src/components/CanvasOverlay.tsx (from fix: correct Fabric text box resizing #984 on main) — untouched.

Needs human verification

• Drag feel at the 600-image scope on real hardware (scripted drags cannot prove feel — process gate in docs/drag-and-drop.md).
• Whether the remaining ~370 ms activation stall at 600 tiles warrants follow-up (deferred/incremental shape measurement).

Link to Devin session: https://app.devin.ai/sessions/414754bd88374dc693f47100ac269e78

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Devin Review found 2 potential issues.

[devin-ai-integration]

Devin Review found 1 new potential issue.