promtool check metrics inconsistencies in "internal" metrics

[awelzel]

Running promtool check metrics for the broker / caf generated Prometheus metrics indicates some conventions are not adhered to.

$ curl -sf http://localhost:4242/metrics | promtool check metrics
broker_processed_elements counter metrics should have "_total" suffix
caf_system_queued_messages_total non-counter metrics should not have "_total" suffix
process_cpu_seconds_total non-counter metrics should not have "_total" suffix

process_cpu_seconds_total could just be switched to a counter. Not sure about the semantics of caf_system_queued_messages_total (gauge vs counter). Might also be just a type difference.

[Neverlord]

The metrics that come from CAF code (the process metrics also come from CAF) probably need to wait for CAF 0.19, the Broker metric I can update right away.