Automatically rescue orphaned background jobs left in executing state

Hive now automatically recovers background jobs whose owning node has disappeared. Previously, a scheduled singleton worker such as domain evolution could get stuck in the executing state forever if the pod running it went away, blocking every future scheduled run and silencing health check-ins. This release adds Oban’s Lifeline plugin with a 30 minute rescue window, so orphaned executing jobs are revived and singleton scheduling keeps working. Routine active jobs keep their normal singleton behavior.