Include 'urgency' in Rails request metrics that result in an ETag cache hit

Follow-up from gitlab-org/gitlab!81303 (comment 854218746) and #1305 (closed).

In gitlab-org/gitlab!81303 (merged) we're adding the correct caller_id to the ApplicationContext, which results in in being picked up in the RequestsRackMiddleware. We've done this by linking a controller and action to the ETag route.

We intended to add the urgency as well, but that was not a simple fix: for the urgency, we require a Struct that has a name (for labelling) and a duration (for comparing the total duration against).

The urgency is currently found in the env hash for the request, where we dig up the api.endpoint or the action_controller.instance to find the urgency.

We should still add the urgency so we consistently report the same urgency for an endpoint, without this, all urgencies will be the default one (1s).