Drop of collection only containing prepares does not schedule compaction

Been looking at MB-41165 and spotted this. I think that this was introduced in MB-41882 which assumes that we use some other mechanism to clean up the prepares. The compaction is required to clean up the HashTable and DurabilityMonitors though. I don't think this causes any issues in terms of the DurabilityMonitors/assertions. They should deal with having dropped but not erased writes lying around already. The next compaction will eventually clean up these prepares but they could be a significant amount of data in certain circumstances so we need should schedule the compaction.