Distributed deadlock detection improvement - Add some server logs for the users

As we've briefly discussed in #1529 on several comments, the logging for distributed deadlock detection should be improved.

We currently have `citus.log_distributed_deadlock_detection` flag, which is mostly intended for developers, not the users. We should expand (or replace) that with some more useful logs for the users.

(Note that the guc is `GUC_NO_SHOW_ALL` thus not exposed to the users)

We should probably check for what Postgres logs when it finds a local deadlock.