test(mcp): run isolated scripts in parallel - #732

Summary

MCP tests run 22 independent Node scripts serially. Use Node's built-in test runner to run two isolated files at a time and report each file's result and duration.

Changes

• Discover every tests/test_*.mjs file instead of maintaining a shell command list.
• Keep assertions within each file sequential and preserve the existing process, SQLite, filesystem, and stdio tests.
• Stream Node's TAP output and write its JUnit report best-effort. Report-directory and file failures warn without changing the test result.
• Upload the MCP report from its actual directory, ignore generated reports, and document serial reproduction.
• Merge current main, including the workflow test split and Windows test fixes. The re-audit found no further simplification needed in the native runner.

Testing

• All 23 MCP scripts passed on Node 22.13.0 and Node 24.15.0. The report contains 23 test cases and no failures.
• On the refreshed branch, 312 focused SDK checks passed, including publication process groups, review transport, workflow configuration, and release safeguards.
• SDK and MCP type checks, SDK formatting, formatting of the changed MCP files, and git diff --check passed.
• The reporter regression covers writable and blocked destinations and preserves passing and failing test results.

Risk and rollout

No production code or existing assertions change. Files run in separate processes with bounded concurrency, and each file owns its temporary state. A failing file still fails the command. Reporting remains optional. Hosted timings should be checked before increasing concurrency further.

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.