Use an Artemis last-value queue for watchStateChanges

[agarciadom]

To reduce the potential overhead of watchStateChanges on a server, we use temporary queues, which normally don't touch the disk at all and do not persist reconnections. However, a client with a very slow or unstable (but not dropped) connection could still have messages piling up in their queue, which would not be useful, as we would only be interested in the last message anyway.

Artemis provides "last-value queues" precisely for this purpose, and we should use them:

https://activemq.apache.org/artemis/docs/1.0.0/last-value-queues.html