query monitoring improvements: vitals keyspace and vital logging

The query monitoring infrastructure includes an endpoint that reports query node vitals.
The issue is, the endpoint needs to be accessed directly and on each node, which makes the vitals not very useful.

Create a vitals system keyspace, so that vitals for all nodes can be accessed through n1ql queries.

Also, implement vitals logging at specific intervals, so that support can have some sort of heartbeat and be able to follow resource usage over time.

In both cases, information returned should be
total requests processed and rejected
current active and queued requests
requests accepted, serviced and rejected since the last interval
current and max memory usage
some sort of cpu and load metric (max and in the last interval)