feat: CI failure auto-diagnosis for evolution PRs (Closes #577)

[Da-Mikey]

Automated evolution PR for issue #577. Adds evolution_ci_diagnosis.py — polls open PRs on the Da-Mikey fork, fetches failing CI logs, extracts error classes, and either auto-fixes trivial failures (lint, unused imports, trailing whitespace) or creates focused child issues for complex failures. Runs on a 30-minute cron schedule.

[Lexus2016]

Integration review: blocked from merge. The implementation uses "gh pr view --json recentCheckRuns", but the installed gh CLI (and the version available in the cron environment) does not expose a "recentCheckRuns" JSON field, so the script cannot retrieve CI status and will silently do nothing. This is a category error: the mechanism cannot deliver the issue's goal (auto-diagnose failing CI) as written.

Rework brief:

• Replace fetch_latest_ci_run() and fetch_failing_log() with supported gh/REST approaches: parse "gh pr checks --repo " output, or call the GitHub REST API "repos///commits//check-runs" / "repos///actions/runs".
• Add a test that exercises the CI-status fetch with a mock runner returning realistic check-run JSON, proving the function actually extracts a failure.
• Verify in the PR description the exact gh version/fields the code relies on.

Re-open a PR once the script can reliably detect a failed CI run for a real open PR.