Add more Prometheus metrics

#### Version information:

0.4.17

#### Type:

Enhancement

#### Description:

Infura is hosting a set of IPFS nodes to provide a public IPFS API. We would like to monitor more metrics not currently exposed by go-ipfs as far as I can tell:

- bandwidth in/out (as given by `ipfs stats bw`), potentially split for the various subsystem (DHT, bitswap ...)
- repository size and object count (as given by `ipfs repo stat`)
- peers latency distribution (processed from `ipfs swarm peers --latency`)
- DHT metrics (latency, failure rate ...)

Is that something that would interest you ? I can work on that, but a few pointers would be appreciated.

CC @lgierth