semaphore handling could use improvement

[Paul Collins]

If the sync is killed, it can end up in the following state:

2020-05-05 20:02:01,240 - Can't acquire semaphore: /tmp/metadata-sync-running already exists.

This file should at least be a pid file, with a check that the pid exists and is a python3 process, or suchlike, and if not, then we force the lock.

However, we already wrap the cron invocation in run-one, so this seems spurious.

[Haw Loeung (hloeung)]

This is probably not needed since the cron job to fire the sync is wrapped by run-one.