journal_client: Tag 'get_origin_metadata' and 'metadata_fetchers' with forge type + Add metrics for the number of metadata items

There are two commits in this MR:

1. journal_client: Tag 'get_origin_metadata' and 'metadata_fetchers' with forge type

These values vary across forge types (and instances), so it makes sense to be able to query the runtime of 'get_origin_metadata' independently to see if a particular forge is slow.

2. Add metrics for the number of metadata items

existing_metadata_items is capped to 1, at least for now, because it would be wasteful to fetch more metadata just to publish a metric which can probably be computed more efficiently elsewhere if needed.

This might help with swh/meta#5035 (closed) In particular, I'd like to know why the number of ops/s is so much higher than the number of Github requests and the lag derivative.