Adds telemetry on number of DNS queries served, per-agent.

[slackpad]

Adds a counter increment for each successful DNS query (reverse or forward). We had a request to aggregate this per-agent so I suffixed with the node name. We haven't done metrics per node before, though we do a similar thing for services, and Raft metrics roll up per-IP, so there seemed like precedent.

[ryanuber]

This seems reasonably useful. We should probably tack this into the deferred function in each handler, unless there is a reason we are trying to avoid logging the metric when doing recursion. Do you think it would also be worth it to log a metric for the duration since we have the timestamp handy? I could see that being useful for the recursion case to spot problems downstream.

[slackpad]

@ryanuber good call. We get a "count" metric for free with the gauge so no need for two metrics.

[scalp42]

Still hoping we get to have metrics without the hostname in the actual metric name 😇

[slackpad]

@scalp42 ah yeah this will make that a little worse temporarily. I don't think this'll make it into the next point release but I'll try to get #1241 in the release after that.