Garbage collection

[cavis]

Cleanup our stuff.

A bunch of our data is paranoid now. But we don't have any process/hooks/crons to actually delete S3 files.

• Write some async jobs to actually go delete S3 resources. MediaResources/Images should delete their files, maybe episodes should rm -rf their S3 directories, maybe podcasts as well?
• Wire those jobs into the after_real_destroy callback, if that's not too dangerous.
• Also - we can reap Tasks more often. Write a cron that goes and looks for tasks that aren't the latest for their owner, belong to soft-deleted owners, etc.
• Make sure owner-destroys cascade to the Task

Some open questions around how long we retain data for old shows. But should figure that out as part of this ticket. And then:

• Go through our oldest podcast_ids and delete that we don't care about.
• Maybe we need this in staging, for integration tests, as well

[cavis]

Also wondering - should BigQuery still have a record of really-deleted podcasts?

Right now dt_downloads and dt_impressions will forever have those. But podcasts / episodes / etc get overwritten prett frequently, so the gone-show would disappear there.