In some scenarios with a lot of concurrent DDL we want to ensure checkMetadataConsistency eventually gets the lock in order to do the checks. For this, we propose to add a new flag (it could be called something like forceMetadataLockAfterBackoff) that will aggresively grab the lock if it is unable to grab it after some some time.

Also as part of this ticket, we should change how the ddl lock manager is using the backoff strategy, it should by default give up if it doesn't manage to grab the lock, and, if specified, instead of giving up, forcefully grabbing it, like it is working today.