Whenever a parent-child relationship describes a specific path but the
entry's actual path is different, this new repair step will adjust it.

In the event where another entry already exists under that name, the
latter is deleted and replaced by the above entry. This should help
preserve the metadata associated with the original entry.

When fixing failed cross-storage moves in the file cache and using the
storage id filter, we filter by target storage for phase 1. However, we
also need to fix the source storages in phase 2. To do so, a list of
affected source storages is now gathered in phase 1 to be run on phase
2.

This instead of iterating over all storages which is way less efficient
due to the 1-N nature of potential failed cross-storage moves that we
are repairing.

If singleuser mode is enabled and "--all --repair" is passed, all
storages will be repaired in bulk (no repair filter). If not, it will
fall back to iterating over each storage which is slower.

…or isolation