Monitoring methods for pure gRPC backends

I am currently trying out a pure gRPC backend with ESP v2 in front of it. The backend is deployed in Cloud Run. Client is a pure gRPC client, meaning there is no expectation that REST <-> gRPC transcoding would happen.

Once deployed, even though it works as it intended functionally, the ESP v2 doesn't log the per-method stats in monitoring. I'll try to define the HTTP mapping to see if that helps, but generally, it doesn't sound like it should be required for this to work on pure gRPC.

Reproduce the pure gRPC deployment described with ESP v2 in front of a Cloud Run backend, using a pure gRPC client. Inspect the monitoring output and compare it with a deployment using an HTTP mapping. Done means per-method monitoring statistics are emitted without requiring REST-to-gRPC transcoding.

Written by the indexing model from the issue text.