[knot-resolver 6] Prometheus metrics improvements.

[Jean-Daniel]

Hello,

I'd like to suggest a way to improve prometheus metrics.
I think it would be better to export specific metrics using labels, instead of creating one series per tag;
For instance:

resolver_request_total. Instead of having one metric per kind, resolver_request_udp_total, resolver_request_tcp_total, …
it would be more practical to get a single resolver_request_total metric with a kind label.

It makes it easier to query prometheus and create graphs in Grafana. You can create a single graph in Grafana with a single request grouped by kind instead of having to known every kind in advance, and create a query per kind.

For instance: sum(rate(resolver_request_total[$__rate_interval])) by (kind) will display on line per kind, and if a new kind is released (doq), no need to update the Grafana dashboard.

Ditto for resolver_answer_total and rcode. Having resolver_answer_total{rcode=""} would make querying easier.