- add a priority attribute to the cluster nodes
- when getting peers needed to handle a query, use the nodes
  with the lowest priority.
- update kafkamdm to track kafka lag, and use this lag as the
  node priority.  So if a node has a large lag (either due to
  proecessing the backlog at startup, or because of a fault) it
  will stop recieving queries.  But if all nodes in the cluster
  have high lag, then queries will still be processed and the user
  will get all of the data available.
- closes #519

- instead of just using the number of messages the node is lagging
by, divide the number of messages by the ingest rate to get the
number of seconds of lag.  ie, if a node is ingesting at 10k/s
and the lag is 1k, then the priority is set to 0, as the outstanding
metrics will be processed within 1second.  But if the ingest rate
is 10k, and the lag is 40k then the prio gets set to 4 as it will take
4seconds to process the outstanding metrics at the current rate.

… (lower priority)

For kafkamdm specifically this allows us to remove a node from service if it is lagging
by too much (measured in seconds of lag)