Add Prometheus Metrics

[arlyon]

This PR publishes information about the pusher and fetcher to the prometheus metrics endpoint.

What is left out of this is publishing qualifiers, which may be an interesting metric, but will be included in a subsequent PR

[tomcoldrick-ct]

Nice one, thanks a lot! Overall I think this patch looks great, just a few nits (plus responses to your comments and some todos).

I have written an impl for metrics_asset_store but haven't hooked it up yet. Will we need that? From what I can tell publishing asset store metrics is somewhat covered by the others.

In my humble opinion this is adequately covered from the blobaccess metrics along with the Fetch and Push server metrics.

I have published duration_seconds and size_bytes. I believe most other interesting things can be inferred. Is that thinking correct?

I think qualifiers may be an additional vector we might want, given their flexibility. From an operator perspective I'd probably like to know if, e.g. my git qualified fetches are failing more often or whatever. That said, such a thing is probably obvious anyway. Another potential thing that crossed my mind while looking at this is some kind of metric to allow me to calculate "% cache hit", however in the current design this is kinda awkward. I suspect that redesign related to #3 will be informative on this, so it's probably best to wait for that in any case.

[tomcoldrick-ct]

Aside from the failing CI (looks like buildifier is not appeased), this looks good to me!