Major Description
While investigating a dead-lock in https://github.com/linagora/james-project/pull/4322 I realized the following:
Spring relies on a distinct JVM path locker for the maildir store,
distinct from the mailbox API level one.
Only tests enforces the use of a common path locker, which do not
reflect production usage.
Note that using a single pathLocker leads to deadlocks as acquiring
the lock at the maildir level is sometime done with a lock already acquired
at the mailbox level.
Fixed a part of https://github.com/linagora/james-project/pull/4322