WAL-E container not confirmed empty before using

[Stuart Bishop]

If redeploys are made using the same WAL-E target URL, the container or bucket is used as is. If it contains cruft from previous deployments, then the old WAL and history files will cause problems when secondary databases get restarted.

Solution is to confirm the bucket is empty before using it, going into a blocked state if it already contains data. We don't want to delete any existing data as it might be important. The leader could do this, making a note if it is clean in leadership settings. Until this flag is set, units should not configure PITR and be in a blocked or waiting state.