Add rabbitmq_endpoint label to rabbitmq_identity_info (backport #13218) (backport #13237)

[mergify]

This allows to differentiate between metrics with the same name, coming from the /metrics/ endpoint and those coming from /metrics/per-object (or from /metrics with prometheus.return_per_object_metrics = true).

The lack of such differentiation can easily lead (and leads in our Grafana dashboard) to some values being doubled if both endpoints are scraped (which is a reasonable thing to do - collect aggregated metrics often for a high-level overview and per-object metrics less often not to overload the system). This leads to metric series like these:

# aggregated metric
rabbitmq_queue_messages_ready 53
# per-object metrics
rabbitmq_queue_messages_ready{vhost="/",queue="qq-2"} 20
rabbitmq_queue_messages_ready{vhost="/",queue="qq-3"} 20
rabbitmq_queue_messages_ready{vhost="/",queue="qq-1"} 13

a PromQL query such as sum(rabbitmq_queue_messages_ready) returns 106, even though there are only 53 messages ready. With this change, you can write a query such as rabbitmq_queue_messages_ready * on (instance, job) rabbitmq_identity_info{rabbitmq_endpoint="aggregated"} so that per-object values are ignored

---

This is an automatic backport of pull request #13218 done by Mergify.

---

This is an automatic backport of pull request #13237 done by Mergify.