Parse unavailable indexer health messages and emit metric

[rtrox]

Description of the change

Prowlarr doesn't consider indexers that it's marked unavailable due to errors as "disabled", it only considers them "unavailable". This unavailable status doesn't have a metric of it's own, it just shows up as a health message, with the server names in a comma delimited list in the message.

This change looks for IndexerStatusCheck health messages, and emits an extra metric per indexer.

Benefits

This extra metric will make it much simpler to build alarms, as one can key off the number of indexers unavailable (or even % of indexers unavailable by dividing this by indexers_total or indexers_enabled)

Possible drawbacks

Applicable issues

• fixes Parse Indexer Unavailable Errors to emit unavailable_indexers metric #158

Additional information

Metrics:

# HELP prowlarr_indexer_unavailable Indexers marked unavailable due to repeated errors
# TYPE prowlarr_indexer_unavailable gauge
prowlarr_indexer_unavailable{indexer="ServerOne",url="https://my.instance"} 1
prowlarr_indexer_unavailable{indexer="ServerTwo",url="https://my.instance"} 1
prowlarr_indexer_unavailable{indexer="ServerThree",url="https://my.instance"} 1
prowlarr_indexer_unavailable{indexer="Server.four",url="https://my.instance"} 1