Add Executor metrics for its total,free,used logical resources (absolute and possibly % utiliazation numbers)

We only have such metrics for physical resources. This doesn't allow us to troubleshoot capacity problems in clusters. Like answering questions "are all executors currently maxed out from Server perspective?"
