Paused jobs can be easily forgotten about. Long-paused jobs can cause cluster issues, because of the accumulation of garbage that paused jobs can create. Additionally, supporting long-paused jobs through cluster upgrades is hard.

We should:

• Expire paused jobs after 3 days of being paused, add expiration time as a column in jobs table
• If a job is unpaused then paused, the expiration counter should reset

Rational behind 3 days: you have enough working-day time to have done something about this

Follow up work:

• Make the default pause limit configurable (jobs: add configurability to paused jobs expiration #84600)
• Display expiration time in DB Console (jobs: display job expiration time in the DB console #84599)
• Add the ability to extend this time jobs: follow up paused jobs work (extend time) #85978

Jira issue: CRDB-17756

Epic CRDB-20422