jobs: improve observability around whether a job is in an idle state or not

Some background, forever-running jobs always mark themselves as idle so as to never prevent the SQL pod from being scaled down in our serverless offering. Jobs like the auto-config runner and GC jobs mark themselves as idle during certain phases of execution when they are not actively executing, and not idle during others. Currently, the idle state of a job is only stored in memory in the list of jobs that a node registry has adopted. It would be useful to persist this information in the `job_info` table and surface it via SHOW JOBS and in the DB console.

Jira issue: CRDB-27001