Prometheus metrics policy

APIcast policy which allows custom metrics to be created and exposed at the "/metrics" endpoint for collection be a Prometheus metrics server.

A default set of metrics will cover the most common use cases and set an example on how custom metrics are defined:

• 3scale-auth status codes: Total, 2xx, 4xx, 5xx
• Upstream status codes: Total, 2xx, 4xx, 5xx
• Connections (per Joaquim): Read, Write, Wait, Open
• Request time:
• - must full end to end latency
• - must upstream latency
• - optional breakdown of latency in APIcast pre- and post-request

Linked Github issue:
https://github.com/3scale/apicast/issues/745