perf(ai): batch chat stream updates

[JustYannicc]

Summary

• Batch AI chat streaming message visibility through a 40ms stream buffer instead of updating React state for every chunk.
• Keep the authoritative assistant response in the buffer and force a flush before completion, error handling, cancellation, and mode transitions.
• Add a focused stream batching harness that records update/render-work counts and covers scheduled streaming, final completion flush, and error flush persistence.

Metrics

Focused harness: node scripts/test-ai-chat-stream-batching.mjs

Mode	Chunks	Visible updates	Layout measurements	Simulated markdown reparse chars	Elapsed
Unbatched baseline	240	240	240	267,795	0.055ms
Batched burst	240	1	1	2,290	0.094ms

Baseline was first captured before the production hook change with 240 visible updates for 240 chunks.

Validation

• node scripts/test-ai-chat-stream-batching.mjs passed.
• node --test 'scripts/test-*.mjs' passed: 25 pass, 1 skipped live Electron test.
• ./node_modules/.bin/tsc --noEmit --target ES2020 --module ESNext --moduleResolution bundler --jsx react-jsx --strict --esModuleInterop --skipLibCheck src/renderer/src/hooks/useAiChat.ts src/renderer/src/utils/ai-chat-stream-buffer.ts passed.
• ./node_modules/.bin/vite build passed.
• node scripts/check-i18n.mjs completed; it still reports existing Italian locale gaps for settings.ai.whisper.vocabulary and settings.extensions.appSearchScope.

Notes

• Whole-repo tsc --noEmit still fails on existing unrelated diagnostics outside this change.
• Local pnpm test was blocked by pnpm build-script approval; the underlying package test command was run directly with node --test 'scripts/test-*.mjs'.

[JustYannicc]

Consolidated into #531, which combines the AI chat streaming, Cursor Prompt streaming, Raycast useAI streaming, and AI status polling performance work in one upstream PR. Closing this superseded scoped PR in favor of the consolidated replacement.