This is a draft, not ready to be merged. I wait with test cases until
getting feedback about the main functionality.

Same functionality as in buchgr#350

 - Prometheus counter for cache hit ratio of only AC requests.

 - Support for prometheus labels based on custom HTTP and gRPC headers.

but implemented in an alternative way:

 - disk.io implements two interfaces: cache.CasAcCache, cache.Stats

 - cache.metricsdecorator is decorator for cache.CasAcCache and
   provides prometheus metrics.

Pros with this alternative implementation:

 - Should allow non-prometheus metrics as requested in
   buchgr#355

 - Avoid the question about if the counters should be placed in
   disk.go or http.go/grpc*.go. If placing in disk.go, there are
   issues about how to avoid incrementing counter twice for the
   same request (both in Get and in GetValidatedActionResult)
   and at the same time count found AC but missing CAS, as cache miss.

 - Makes headers available also for logic in disk.go, which could be
   useful for other functionality in the future.

 - Metrics can be separated from logging, and still not require
   injecting counter increment invocations in tons of places.
   Incrementing from a few well defined places minimize the risk
   for bugs in the metrics.