[FEA] Delete cached Spark event logs when they are no longer needed #281

Is your feature request related to a problem? Please describe.

The tool saves event logs so a timed-out attempt can pick up where it left off. Once it has finished calculating a job's cost, those saved logs usually serve no further purpose, but they stay on disk.

Describe the solution you'd like

Keep the files while a job still needs retries, then remove them after a successful calculation. Do not remove a file that another calculation is using. Test retries, successful cleanup, and calculating the same job again later.

Describe alternatives you've considered

Operators can clear the cache while callers are stopped, but that also removes useful progress for other jobs still awaiting retries.

Additional context

Follow-up to #278 and the discussion about event-log cache lifetime: #278 (comment)