Improve metrics setup and docs

[britto]

While building a custom metrics module for hackney I noticed a few inconsistencies in how some metrics are set up and documented. This PR addresses three of them. I've split them into separate commits to make it easier to review and eventually drop whatever may be irrelevant.

1. Fix the queue_count metric documentation and setup. It was being initialized as queue_counter in metrics:new;
2. Update pool metric docs to match their names. They are currently prefixed with hackney_pool, but show up prefixed with hackney in the docs;
3. Add documentation for 2 host-based pool metrics: new_connection and reuse_connection.

Update: Sorry, just noticed 1. is the same change proposed at #583, 2. and 3. are still valid though.

[benoitc]

thanks!