NRPE check for detecting RabbitMQ split brain

[Paul Goins]

We may want to add a check which would check logs (or elsewhere) to detect whether RabbitMQ is potentially running in a split-brain situation.

We saw a situation on a customer cloud where the cluster looked OK per "rabbitmqctl cluster_status", however messages were being dropped (e.g. "Discarding message {'$gen_call',{<0.1234.0>,#Ref<0.1231231231.4564564564.123456>},stat} from <0.1234.0> to <0.56789.9876> in an old incarnation (1) of this node (2)"). We ultimately needed to shut everything down and rebuild the cluster.

An alert for this would save time in diagnosing the problem.