[codex] Reconcile stale running jobs when pid exits

[LawrenceChiu95]

Summary

• auto-reconcile stale running jobs when the tracked pid no longer exists
• persist the reconciliation to both state index (state.json) and per-job file (jobs/<id>.json)
• append a log line so users can see why a job switched to failed

Root Cause

/codex:status trusted persisted state for running jobs without validating whether the worker process was still alive. If a worker died unexpectedly, the job could remain running indefinitely and appear stuck.

Changes

• plugins/codex/scripts/lib/state.mjs
  • add pid liveness check (process.kill(pid, 0))
  • add reconciliation path in listJobs() for stale running jobs with dead pids
  • mark stale jobs as failed with errorMessage, completedAt, and pid: null
  • mirror updates into per-job json and append a log annotation
• tests/state.test.mjs
  • add regression test: listJobs auto-reconciles stale running jobs when pid is no longer alive

Validation

• node --test tests/state.test.mjs
• node --test tests/runtime.test.mjs --test-name-pattern "status shows phases|status preserves adversarial review kind labels|status --wait times out cleanly|cancel without a job id ignores active jobs from other Claude sessions|cancel with a job id can still target an active job from another Claude session|stop hook logs running tasks to stderr without blocking when the review gate is disabled"