The current implementation of the create collection coordinator does not guarantee a clean exit when receiving a non-retriable error. We are doing a best effort to release the critical section, without any guarantee that it is going to succeed (onError from create_collection_coordinator.cpp).

The goal of this ticket is to add a new clean up procedure on abort that guarantees that before exiting the create collection we will, at least, release the critical section and possibly remove orphaned collections on the participants.

This work is similar to what the move primary coordinator introduced:

• https://github.com/mongodb/mongo/blob/1a513ac4988dc77b74d7e7270b2c6968ee9afe5a/src/mongo/db/s/move_primary_coordinator.cpp#L324-L333
• https://github.com/mongodb/mongo/blob/1a513ac4988dc77b74d7e7270b2c6968ee9afe5a/src/mongo/db/s/move_primary_coordinator.cpp#L351-L404