Add per-Process event queuing delay metrics.

In a similar style to MESOS-9235, we can introduce per-Process metrics that capture the trip time through event queues on a per-Process basis. This could be used to get a better sense of how much of a backlog a particular actor has (e.g. it's taking 10 seconds for an event enqueued into the master to be dequeued).

Ideally this would be of a histogram format (e.g. expose percentiles over a window), but a simple start is to expose the latest time measurement.