Improve latency of api/:version/repos/:namespace/:project/branches

The latency >1s count of api/:version/repos/:namespace/:project/branches is currently the highest in that table, and is likely to be contributing to the ~"group::ecosystem" error budget cost.

We should try to improve the latency of this endpoint.

Note, the latency issue could possibly relate to #331329 (closed), but more likely doesn't.

The endpoint has been given a temporary custom Apdex SLO of urgency: :low (5 seconds, rather than the normal 1s threshold) in !78142 (merged) until we have optimised this endpoint. Part of this issue is to remove the custom Apdex SLO for the endpoint when we have verified that performance has been improved.

Testing & Availability

Suggestion: Consider adding a new scenario in the GitLab Performance Tool to cover this endpoint.