Add a counter for coordinator changes through the operator

[johscheuer]

This could be helpful to detect if the operator changes very frequently the coordinator, which could either be a bug or cluster instability.