perf(launcher): optimize command list rendering

[JustYannicc]

Summary

• Window large launcher command lists so only the visible row range, overscan, and selected-row neighborhood render.
• Memoize launcher row derivations for labels, icons, hotkeys, refs, and handlers with stable parent callbacks.
• Add a focused render measurement harness plus a regression test that fails if 5,000-command scenarios render every row again.

Before / After Metrics

Measured with node scripts/measure-launcher-command-list-render.mjs --rows=5000 --iterations=5 --warmups=1.

Scenario	Before row renders / median	After row renders / median
root results initial	5,000 / 123.38ms	18 / 1.35ms
selection moves to middle	5,000 / 126.65ms	33 / 1.57ms
selection moves to end	5,000 / 116.01ms	26 / 1.33ms
query update same-size reshuffle	5,000 / 129.20ms	17 / 0.56ms
query update narrowed-large	3,000 / 85.45ms	28 / 1.17ms
Total scenario sequence	23,000 / 580.69ms	122 / 5.99ms

Validation

• node --test 'scripts/test-*.mjs' passes.
• node --test scripts/test-launcher-command-list-windowing.mjs passes.
• node scripts/check-i18n.mjs completes; existing Italian locale misses settings.ai.whisper.vocabulary and settings.extensions.appSearchScope.
• node /opt/homebrew/lib/node_modules/npm/bin/npm-cli.js run build:renderer passes.
• node /opt/homebrew/lib/node_modules/npm/bin/npm-cli.js run build:main passes.
• Codex LSP diagnostics for LauncherCommandList.tsx and LauncherCommandRow.tsx report no errors.

Notes / Risks

• The virtualizer uses fixed row/header/calculator height estimates matching the current launcher row layout; future row-height changes should update those constants.
• Project-wide ./node_modules/.bin/tsc -p tsconfig.renderer.json is currently blocked by unrelated existing errors outside this change.

[JustYannicc]

Superseded by the consolidated renderer performance PR #533: #533. Closing this scoped source PR so review stays centralized there.