Fix prometheus endpoints

[kickster97]

WHAT is this pull request doing?

Fixes the PrometheusController,
Previously the iterator vhosts from target_vhosts() was misused, resulting in the value being 0 for many metrics.
Now we let vhosts be an array, instead of an iterator. This allows us to iterate over it many times, as we do in register_routes.

Also: Adds a type to the metrics that were missing it, so that datadog is able to scrape everything.

HOW can this pull request be tested?

accumulate some queues, messages, exchanges etc. and check the endpoint at
http://localhost:15672/metrics

[jage]

Possible to add specs for the metrics (i.e. integration style, read the prometheus end-point and verify the numbers)? We have a simple parser in Crystal https://github.com/84codes/prometheus.cr

[viktorerlingsson]

LGTM 👍

Agree with @jage , would be great to verify with a spec or two!

[kickster97]

Seems like the prometheus parser is not used in any of our projects, has no starts of forks.. @jage is the intention to maintain it?
A solution could be to just lift in the parser code into lavinmq (its not that many lines of code), as i'm having trouble making it work as a dependency..

[jage]

Seems like the prometheus parser is not used in any of our projects, has no starts of forks.. @jage is the intention to maintain it?

No, provided as-is. There's another non-public parser available, but not sure what's the best fit here.

[viktorerlingsson]

LGTM!

(but run crystal tool format)