dandi cleanup (or gc)

#813 "inspired" to file this issue: we are accumulating considerable amount of logs from `dandi`. E.g. on my work laptop I have 800 of logs. Might not be critical for me, but still consume inodes and space so we might want to add easy way to cleanup after various dandi operations, and have some default `--logs-expiration-age=31` (days; name could be different). Besides logs we might want to

- cleanup fscacher's dandi cache
- might want extra options after https://github.com/joblib/joblib/pull/1200 is finalized/released
- we could even interface/call `datalad clean` having detected being in a datalad dandiset
- .... anything else ... ?

WDYT @jwodder ?