[RFE] Generate metrics

[ingvagabund]

Even the node-fencing controller is another source of interesting activity in the cluster. For that, we should generate various metrics such as:

• how many times a given node was restarted
• how many times a node bad behaviour was caused by a hardware failure
• how many times a squirrel attack caused a network failure
• ...

So operators can see which nodes are sensitive to failures and possible track which failures are happening regularly.