[metrics] Add RID metrics: subscriptions and isa count

[the-glu]

This PR add two RID metrics: subscriptions and isa count

Contrib to #1257

[mickmis]

Having a metric actively pulled from somewhere external to the runtime and doing a DB query sounds a bit dangerous. What about decoupling the DB polling from the the observe callback? And have a routine periodically querying those stats. We already have some form of 'DB' ping that could be put in common.

[the-glu]

Having a metric actively pulled from somewhere external to the runtime and doing a DB query sounds a bit dangerous. What about decoupling the DB polling from the the observe callback? And have a routine periodically querying those stats. We already have some form of 'DB' ping that could be put in common.

I don't think that an issue? It's very simple, readonly calls, using index / simple count only (and probably cached), and this allows calls to fetch as fast or as slow as needed in term of update interval from the monitoring system, without enforcing an update interval.

Causing performances issues would mean DDOS on purpose, on an private, separate endpoint, and this can already be done on others, public by design.

[the-glu]

I added a small cache (1s, not configurable) who should prevent unexpected cases.

[mickmis]

LGTM modulo naming to follow semantic conventions.