As a user, I want a more reliable Orchestrator job model cache.

[lm-sec]

AC:

• Deleting a job from the Orchestrator cache is not currently possible
  • Watch out for the race condition of a job model being deleted while a job of that model is queued
• Ensuring that a job in the cache is no older than the job being ran with an updatedAt timestamp
  • Mimic versioning of jobs with an at least as recent as X mechanic
  • Enforce the promise of running the job in at least the version at the moment of the job launch
• Retry/Close jobs by setting them as failed if they can't find their job model in the cache