extend, document and test metrics support

[pohly]

We already have a metrics service for the controller in our deployment, but so far we are mostly just using it internally during testing to verify that the driver is ready. We also get metrics support from the Kubernetes-CSI sidecars

We should

• check what other information besides "version" and "number of nodes" may be useful
• enable metrics support in the sidecar deployments
• document how metrics support works, including an example for gathering it with tools like Prometheus
• last but not least, add more tests to test/e2e/metrics

[pohly]

Let's target for 0.8.x as part of making PMEM-CSI more admin-friendly.

[pohly]

enable metrics support in the sidecar deployments

Upstream is tracking documentation efforts for that in kubernetes-csi/docs#339 after I asked how to use this.

While trying to enable it, I ran into prometheus/prometheus#3756 which IMHO prevents scraping the sidecar metrics. I tried the approach of merging in their data, but because some of it is the same in all processes (like process_open_fds or go_gc_duration_seconds), that doesn't work properly. Code is in https://github.com/pohly/pmem-CSI/commits/sidecar-metrics

[pohly]

This approach is more promising: https://github.com/pohly/pmem-CSI/commits/metrics-enhancements

[pohly]

Done in PR #715