expose per-endpoint stats in admin interface

[olix0r]

Finagle can (optionally) record per-endpoint statistics. We should add a router-client-level configuration parameter to enable recording of these metrics, and integrate them into the admin UI.

These metrics should be disabled by default because they can be quite costly performance-wise (esp for large clusters), but it's a great feature to have when individual backends matter. In larger clusters, tracing is probably a better way to attack this, anyway.

[olix0r]

We could extend the telemeter API to support allow metrics.