feat: single RMQ node disk alarm -> stop node

If a single node experiences a disk alarm, RabbitMQ will block all publishing connections, making the cluster unusable. This should only happen if a bug causes the node to use up too much disk space, since you usually would expect a disk alarm to occur on all nodes - but if it happens this failover mechanism inside the readiness probe will prevent blocked connections. Before the node is stopped, we ensure that there are no disk alarms on other nodes and that stopping the node does not cause a quorum loss.