perf(files): avoid interval file-index rebuilds

[JustYannicc]

Summary

• Make interval refresh use the existing watcher/incremental path when the index is already built and the watcher is healthy.
• Keep full rebuilds for startup, watcher-error recovery, watcher-unavailable interval fallback, and a 6-hour safety rebuild window.
• Add a focused temp-directory regression/measurement test covering create, update, delete, nested directory indexing, healthy interval ticks, and fallback rebuilds.

Metrics

Baseline before production edit:

• Command: SUPERCMD_FILE_INDEX_BASELINE=1 node --test scripts/test-file-search-index.mjs
• Fixture: 1,043 indexed entries under a temp home directory
• Healthy interval ticks: 3 full rebuilds, 88.38 ms total
• Watcher-style batches: 0 full rebuilds
• Fallback interval with watcher unavailable: 1 full rebuild, 28.15 ms

After this change:

• Command: node --test scripts/test-file-search-index.mjs
• Fixture: 1,043 indexed entries under a temp home directory
• Healthy interval ticks: 0 full rebuilds, 0.06 ms total
• Watcher-style batches: 0 full rebuilds
• Watcher-error recovery: 1 full rebuild, 32.57 ms
• Fallback interval with watcher unavailable: 1 full rebuild, 36.14 ms

Validation

• node --test scripts/test-file-search-index.mjs
• node --test 'scripts/test-*.mjs' (25 pass, 1 skipped live Electron test)
• pnpm run build:main
• pnpm run build:renderer
• pnpm run check:i18n (completed; existing Italian locale warnings remain: settings.ai.whisper.vocabulary, settings.extensions.appSearchScope)
• Codex LSP diagnostics for src/main/file-search-index.ts and scripts/test-file-search-index.mjs: no errors

Risk

If the platform watcher misses an event while still reporting healthy, the index can remain stale until a later watcher event or the 6-hour safety rebuild. Watcher errors and unavailable watcher state still force full rebuild recovery.

[JustYannicc]

Superseded by the consolidated upstream PR #535: #535