In order to mitigate the possibility of the defected interleaving described in SERVER-70600, in case the database already exists we now enforce a clean-up for the primary and secondaries of the related database info (including the dbVersion) before passing to the second phase and flush the cache.

However, there is still a small chance for the primary to recover the dbVersion if there is an stepdown during the execution of the 1st phase, as described in SERVER-70600. Therefore SERVER-70600 won't be completely fixed until we will introduce the Recoverable critical section (SERVER-73391)