Add stats for

• gauge for queue length (waiting to execute)
• work
• work with exclusive
• network requests
• breakdown by network commands for the following
  • queued
  • waiting on network
  • waiting on callback (since they are serialized)
• total counters
  • callbacks
  • schedules/with-exclusive
  • network calls
  • network errors

Do we want to keep track of latency per host/command? If so probably best to add a new issue.

We can add a server metric section or subsection for this, and make it non-default.