Scheduled compaction job should not perform a system.jobs table scan to verify
compaction job does not exist. Doing so is not needed since scheduling system already
does that, and furthermore, it's extremely disrupting to the scheduling system.

Performing system.jobs scan as part of starting the job results in any existing job
(not necessarily started by schedule) causing the schedule to be restarted. Restarting the schedule
results in new job records being created, and increase in the intents, with subsequent lock ups of the
jobs (and schedules) tables.s

Jira issue: CRDB-14143