Add ability to log docker containers/activities

probably should be thought of in the context of
- #31

Two aspects:
- **which processes**: as we know we cannot track docker "things" via session id, so might be needed to just be "aggressive" and log all new appeared images, started (and stopped?) containers?
- how do we log/see CPU / memory consumption of containers?
- should we consider every container to be a single "pid" or do inspect every "pid" of every container? (the latter might be more consistent, but we then need to add tracking not only by session_id, but also by "env_id" of some kind)
- **what to log**: besides those activities *on* images/containers, desired to store `docker log` outputs for them!
- tricky: they might disappear, be removed (`docker run --rm`), so we might need to be "proactive" and as soon as we detect a new container, start `docker log --follow` into a file (it would die when container stops)

Use cases:
- "regular" execution of bids-apps
- current use case was also "monitoring" running of dandi-cli tests to troubleshoot https://github.com/dandi/dandi-cli/issues/1488 . I did spot celery process being busy, but logs seems contain nothing. But may be some other container some other logs where informative/busy?