Gateways must provide some observability statistics

This is one of the initial [gateway project requirements](https://pad.ceph.com/p/rbd_nvmeof_requirements)

The initial requirements page calls out statistics on # of connected clients to a gateway and # of volumes actively attached (per gateway and aggregate). It suggests aggregate IO statistics as well.

Ceph already uses Prometheus, which seems like a fine choice for this.

Stats from each gateway and each discovery controller can collectively paint a pretty complete picture of what hosts are successfully connected to the cluster via the gateway. It's probably possible to confirm that all the hosts expected to discover and connect to the cluster automatically have actually been able to do so, and that the performance guarantees are met for each of them.

Start with the linked gateway project requirements, then inspect how the gateway and discovery controller currently expose or collect information. Define the Prometheus statistics needed for connected clients, attached volumes, and aggregate I/O per gateway and in total. Done means the required gateway and controller metrics are available for confirming connectivity and performance.

Written by the indexing model from the issue text.