Provide mechanism that allows repeating errors to be handled sensibly

MXS-443 and MXS-810 are examples of the log being flooded because of MaxScale logging the same error over and over again.

A mechanism is needed that allows MaxScale or components of MaxScale to handle situations like that in a manner that does not flood the log; by aborting whatever was being done, by retrying with intervals of increasing length, something else.

MXS-823 handles the symptom but there should be mechanisms that allow the flooding to be handled at its core.