feat: runtime polish pass across core/node/native layers

[RtlZeroMemory]

Summary

This PR implements the runtime polish backlog before starting a new ink-compat implementation.

Implemented tasks

1. Layout: fixed nested stack measurement starvation and added regression coverage.
2. Render: clamped stale scroll offsets in virtualized render paths (virtualList/table/tree/file widgets).
3. Widgets: enabled true fuzzy command palette matching (no strict substring gate) + tests.
4. Focus: zone traversal now uses authoritative runtime last-focused map + same-cycle regression.
5. Node backend: ZR_ERR_LIMIT oversized event batches are recoverable (worker + inline), not fatal.
6. Native FFI: hardened js_u64 BigInt parsing to reject negative/overflow values + regression tests.
7. Dim style pipeline: encoded dim attr bit in drawlists, emitted dim/normal intensity transitions in native diff, added tests.
8. Test/CI hardening: reduced e2e profile for macOS/Windows, stronger native smoke checks, profile-driven e2e gating.

Validation approach

• Used paired implementer/validator agent workflow per task.
• Integrated and reran local checks end-to-end.

Local checks run

• npm run lint
• npm run typecheck
• npm run build
• npm test
• npm run test:e2e:reduced
• npm run test:e2e
• npm run test:native:smoke

All above pass locally in this branch.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a37a32c353

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Persist clamped scroll offsets before row hit-testing

This clamps scrollTop only for rendering, but the clamped value is never written back to state, so pointer hit-testing still uses the stale offset (WidgetRenderer computes indices from raw state.scrollTop, e.g. yInContent = state.scrollTop + localY). After data shrink, users can see rows rendered at the top but clicks/context actions map to out-of-range or wrong rows until some other event rewrites scroll state.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Preserve metadata-based command matches when fuzzy sorting

Removing the filterItems(...) step means getFilteredItems now keeps only sortByScore results, and fuzzyScore evaluates label text only. Queries that used to match by description or shortcut (for example, searching by a keybinding string) will now return no results, which is a functional regression in command palette filtering behavior.

Useful? React with 👍 / 👎.