test(runtime): close shell-run stores before cleanup

[kyletser]

Summary

Track every SQLite ShellRun store opened by the shell-run-manager suite and close those leases before removing the temporary workspaces. This removes the deterministic Windows EBUSY cleanup failure without retrying or masking deletion while a live owner remains.

Refs #2624

Verification

• npm --workspace @maka/runtime run typecheck — passed.
• npm --workspace @maka/runtime run build — passed.
• Windows PowerShell UTF-8 gate from windows-baseline.yml — 2 passed, 0 failed, and exited normally.
• Mutation check: temporarily removed the store close() loop; both selected assertions still passed, then the top-level hook failed with EBUSY ... runtime.sqlite and exit code 1. Restoring the loop returned the gate to exit code 0.
• npm run lint and npm run format:check — passed.
• npx knip --workspace apps/desktop and npx knip --workspace packages/ui — passed (desktop reported two existing configuration hints).
• Full shell-run-manager file on Windows — 42 passed, 9 failed, 8 skipped. The remaining failures are the other Windows PTY behavior slices already described in ci(windows): eliminate hidden failures in the non-blocking baseline #2624; the runtime.sqlite cleanup hook failure is absent.
• Root npm run build / npm run typecheck reach Desktop and then fail on the three pre-existing SessionChangedEvent.type errors in goal-services-adapter.test.ts. Current main fails at the same Build step: https://github.com/apache/maka/actions/runs/32652983938

AI use

Select exactly one:

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Codex diagnosed the ownership boundary, implemented the test-lifecycle change, ran the verification and mutation checks, and drafted this pull request description. The commit includes the required Generated-by: Codex trailer.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

The affected Runtime checks, lint, format, and selected Windows gate pass; the unchecked item reflects the current unrelated Desktop build/typecheck baseline failure documented above.

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[kyletser]

Rebased this PR onto current main at a8661ab, which includes #3642 and removes the three pre-existing SessionChangedEvent.type build errors from the initial run. The new workflow run is currently action_required, so it needs maintainer approval before CI can execute. The branch still contains only the original targeted runtime test cleanup commit.

[Astro-Han]

Reviewed at exact head a8661abd. No findings — no [P0]–[P3]. Publishing on behalf of a reviewer without write access here; I re-verified the scope and gate state before posting.

The diff is confined to the runtime shell-run test lifecycle. The wrapper tracks every ClosableShellRunStore it creates, and the top-level after closes each store before removing the directory — which is the right order, and the whole point of the change. Nothing in the production path moves.

Independent validation at this head: focused build plus the shell-run-manager suite at 58 passed / 4 skipped / 0 failed, with lint, format, and git diff --check all clean.

Not approving yet, and the reason is procedural rather than substantive. When this was reviewed the exact-head checks had never executed — the run was parked at action_required, which is the state fork PRs sit in until someone releases them. I have since released it and test is now running. Never-run is not the same as green, so the approval has to wait for a terminal result on this exact head. Assuming it lands green and the head does not move, this is ready.

[Astro-Han]

Gate now satisfied at the same head a8661abd. The run I released has landed: test is terminal completed/success, which is the only check this path set triggers. The head has not moved since the review above, there are no unresolved threads, and no [P0]–[P3] were found.

Approving.

[Astro-Han]

LGTM — merged. Thanks for the contribution!

中文

已合并，感谢贡献。