feat(test): add case-level executor scheduling

[quanru]

Summary

• add case-level scheduling to Midscene Test with configurable concurrency, resource locks, retries, and deterministic case selection
• introduce a generic Test Executor API for delegating isolated cases to local processes, containers, device services, or remote runners without coupling the framework to a specific infrastructure provider
• add CLI selection by file, directory, explicit case ID, and tag
• persist per-case execution metadata and artifacts in the unified test result and HTML report
• expose aiWaitFor as a built-in Midscene Test node
• document the new scheduling and extension APIs in English and Chinese

Reliability and security

• use a strict JSON-safe Executor DTO and validate the complete result and status invariants before hydration
• prevent remote responses from supplying local report paths; reports must be materialized into a task-scoped directory and referenced by opaque IDs
• isolate project-mode and case-mode orchestration while sharing result and progress helpers
• reject unsafe case IDs and attempt IDs before they can be used as result paths
• release resource locks correctly when an Executor throws synchronously
• preserve project-mode behavior as the default for backward compatibility
• document generated IDs as deterministic rather than stable; persistent selectors require explicit IDs

Report support

• include Executor name, resources, attempts, lifecycle, failures, artifacts, and metadata in the HTML manifest
• render Executor details, artifacts, and metadata in the Test report UI

Validation

• pnpm run type-check:tests
• pnpm exec nx test @midscene/test (343 passed)
• pnpm exec nx build @midscene/test
• pnpm exec nx test @midscene/report (148 passed)
• pnpm exec nx build @midscene/report

Pre-release

Published as 1.12.7-beta-20260916022808.0 under the npm beta tag. A refreshed pre-release will be published after review fixes are accepted.

[cloudflare-workers-and-pages]

Deploying midscene with    Cloudflare Pages

Latest commit:	bc3537a
Status:	✅  Deploy successful!
Preview URL:	https://cbff5d84.midscene.pages.dev
Branch Preview URL:	https://codex-review-test-case-execu.midscene.pages.dev

View logs

[quanru]

Addressed the four review findings in bc3537a:

1. Replaced the Executor/internal-runtime boundary with a strict JSON-safe DTO, full Zod validation, status/attempt invariants, and task-scoped opaque report references. Remote data can no longer supply local reportPaths.
2. Added Executor/resources/attempts/failure/lifecycle/artifacts/metadata to the HTML report contract, manifest builder, report UI, and tests.
3. Split project-mode orchestration into project-execution.ts and shared result/progress helpers in execution-result.ts; the main runner is now 597 lines instead of 929.
4. Corrected English and Chinese docs: generated Case IDs are deterministic but not stable across insertion/reordering; persistent selectors must use explicit IDs.

Validation: pnpm run type-check:tests; 343 @midscene/test tests; @midscene/test build; 148 @midscene/report tests; @midscene/report build.