Mark resource group assignment as urgent and CPU bound

What does this MR do and why?

Mark Ci::ResourceGroups::AssignResourceFromResourceGroupWorker as high urgency and CPU bound.

This worker assigns resources from resource groups and enqueues waiting jobs to be executed by runners. Delays in this worker add user-visible latency to jobs waiting on resource_group constraints.

Rationale for CPU bound: Over the last 7 days (https://log.gprd.gitlab.net/app/r/s/DfSmP):

• p50 cpu_s / duration_s tends just over 0.4, meeting the 0.33 threshold.
• p95 cpu_s / duration_s tends around 0.65.

Rationale for high urgency: It blocks pipeline processing for jobs waiting on resource groups, and over the last 7 days (https://log.gprd.gitlab.net/app/r/s/smjWM):

• 7-day 3-hourly p50 duration_s is around 0.15s.
• 7-day 3-hourly p95 duration_s is under 5s.

The pgbouncer-ci pool concerns noted in the work item discussion are resolved (graph).

This intentionally keeps data_consistency :always. The worker writes while assigning resources, and free_resources = resource_group.resources.free.count must be current to avoid allowing too many concurrent CI jobs to start for a resource group.

References

Part of #593064 (closed)

How to set up and validate locally

Run the worker spec:

bundle exec rspec spec/workers/ci/resource_groups/assign_resource_from_resource_group_worker_spec.rb

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.