Monitor worker saturation on each host individually

Single node workers is a similar metric to the normal unicorn worker metric, except that it only measures a single host.

This reason for this is that once HAProxy has routed a request to a single host it is (for now) "trapped" on that host until its complete. Even if extra hosts are added to the fleet, once the request is on a saturated host, it cannot "escape" and will need to queue on that host until it is served (or cancelled)

There are cases where a single host will receive an unbalanced amount of traffic. In these cases, unicorn saturation could occur on that host without happening across the fleet in general.

This metric helps us keep track of single host unicorn saturation, as opposed to fleet-wide unicorn saturation.

An example of this was https://gitlab.com/gitlab-com/gl-infra/infrastructure/issues/7916

How it looks