Crawl disks of Gitaly nodes in the cluster to clean up left over repositories
Gitalys may contain stale repositories as a result of failing RPCs. For example, Gitaly may create a repository as part of an RPC. If the call fails, this stale repository may be left on the disk. Likewise, failed rename operations would have also left orphaned repositories on the disk.
We should have a crawler in Praefect that periodically compares the contents of the Gitalys to what it expects to be present and removes everything that should not be there. This cleans up stale repositories left over from various failed operations and frees up disk space on the Gitalys.