aperture: internally aggregate session statistics to reduce cardinality

[djkazic]

This PR adds aggregation and classification of sessions into {active, standby, inuse} categories and makes direct metrics exportable.

It is an attempt to lower the cardinality of the prior used read count metric. This should greatly reduce the load on Prometheus for both storage and querying.

However, there is a caveat with the code as-is:

• Unbounded memory growth: while the prior approach essentially made it the prometheus server's problem, we now have to deal with this ourselves. Looking for feedback on an appropriate culling threshold. I wanted to first and foremost mimic what we were doing before so I did not implement culling yet.

[djkazic]

Pushing a lint fix.

[lightninglabs-deploy]

@djkazic, remember to re-request review from reviewers when ready

[djkazic]

Force pushed some changes around TTL and formatting / comments. PTAL @Roasbeef

[Roasbeef]

LGTM ❄️

[guggero]

Thanks for looking into this.
Were you able to test this with the LNC test page?
Would perhaps make sense to make sure the numbers will be what we expect. Maybe @ViktorTigerstrom has time to do a test run?

[guggero]

utACK, LGTM 🎉