fix(gc/worker): delete task if dangling manifest no longer exists on database
I noticed that if a manifest is dangling but no longer exists on the DB (should never happen, but I made it happen by manually deleting data on the DB), the manifest worker will simply log a warning and exit. This causes the tasks to be left behind and picked over and over again with the same result.
This MR makes sure that we delete such tasks if this ever happens.