Optimize ApplicationRateLimiter Redis calls

This is similar to gitlab-com/gl-infra/observability/team#2211 (closed)

We currently run a lot of EXPIRE commands for rate limiting:

Screenshot_2025-07-16_at_4.04.26_PM

It's possible that a lot of these come from Gitlab::ApplicationRateLimiter specially now that we have Sidekiq resource usage rate limits.

We can use the same optimization we did for Rack::Attack and apply it to Gitlab::ApplicationRateLimiter