perf: speed up the test suite

[seratch]

This pull request improves test-suite performance while preserving and, in some cases, strengthening existing behavioral coverage.

It updates pytest-xdist scheduling from loadfile to worksteal, consolidates related cases with parameterization, replaces unnecessary wall-clock waits with deterministic test doubles, and runs independent read-only shell probes concurrently. It also captures the two expected Pydantic non-serializable-default warnings locally instead of emitting them during the full test run.

Local results on the same machine included:

• Non-serial full suite: 53.85s → 28.71s, approximately 47% faster. This measurement is load-dependent.
• Blaxel test file: 5.24s → 1.14s.
• Import side-effect tests: 10.60s → 5.55s.
• Read-path shell probe: 1.46s → 0.75s.
• MCP retry tests: eliminated two fixed one-second backoff waits; the full file now runs in 0.14s.
• Test warnings: 2 → 0.

The final verification completed successfully with formatting, linting, type checking, parallel tests, and the serial test pass.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e7fba314c8

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".