fix agents reaping missing k8s jobs

[kkroo]

Thinking Path

• Paperclip orchestrates AI agents as durable company workers.
• The heartbeat service owns agent run lifecycle recovery when a worker process or external job disappears.
• Kubernetes-backed agents persist a backing Job name in heartbeat_runs.external_run_id, but the reaper still treated a missing namespace-list entry as potentially false negative.
• During the production rollout for PR [codex] fix agent workspace submodule preflight #359, a live opencode_k8s Job was deleted while its run still had fresh output, leaving BLO-9818 marked running with no backing Job.
• This pull request uses the persisted Job name for an exact Kubernetes lookup when the namespace-wide list misses a run.
• The benefit is faster, more precise recovery for deployment-killed agent jobs without weakening the older false-positive protections for list-only misses.

What Changed

• Added readAgentJobRunStatusByName() to classify one persisted Kubernetes Job by exact name, including a first-class missing result for 404s.
• Updated reapOrphanedRuns() to immediately fail/release an external-lifecycle run when its persisted Job name is confirmed missing, bypassing the generic silence floor only for that exact 404 case.
• Added regression coverage for a fresh-output opencode_k8s run whose persisted Job is gone after a rollout.

Verification

• pnpm exec vitest run server/src/__tests__/heartbeat-process-recovery.test.ts — file collected successfully, but embedded Postgres tests were skipped on this host because the Postgres init script exited with code 1.
• pnpm --filter @paperclipai/server typecheck
• git diff --check

Risks

• Low risk. The accelerated reap only applies when external_run_id contains a concrete Job name and an exact Kubernetes read returns 404.
• If the Kubernetes API cannot answer the exact lookup, behavior falls back to the existing silence/staleness heuristic.
• A run whose Job was deleted after actually completing but before Paperclip recorded completion will still be classified as job_missing; that matches existing behavior for missing external-lifecycle jobs with no terminal status.

For core feature work, check ROADMAP.md first and discuss it in #dev before opening the PR. Feature PRs that overlap with planned core work may need to be redirected — check the roadmap first. See CONTRIBUTING.md.

Model Used

• OpenAI Codex, GPT-5 coding agent with tool use and repository-local command execution.

Checklist

• I have included a thinking path that traces from project context to this change
• I have specified the model used (with version and capability details)
• I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work
• I have run tests locally and they pass
• I have added or updated tests where applicable
• If this change affects the UI, I have included before/after screenshots
• I have updated relevant documentation to reflect my changes
• I have considered and documented any risks above
• I will address all Greptile and reviewer comments before requesting merge