Optimize Rack::Attack Redis calls

DRI: @jacobvosmaer-gitlab

As discussed in https://gitlab.com/gitlab-com/gl-infra/capacity-planning/-/issues/562, the way we have configured Rack::Attack in GitLab to talk to Redis is complicated and inefficient.

This issue tracks an attempt to make our use of Rack::Attack more efficient.

• Add new backend gitlab-org/gitlab!112593 (merged) behind feature flag
• Experiment with new backend production#8449, also see https://gitlab.com/gitlab-com/gl-infra/capacity-planning/-/issues/562#note_1292182723
• Remove old backend gitlab-org/gitlab!112944 (merged)
• Remove feature flag env vars from k8s-workloads gitlab-com/gl-infra/k8s-workloads/gitlab-com!2609 (merged)

Status 2023-03-14

We have prepared the feature flag removal MR: gitlab-com/gl-infra/k8s-workloads/gitlab-com!2609 (merged). Once that is merged we can close this issue.

• production#8449 (comment 1291451616)
• https://gitlab.com/gitlab-com/gl-infra/capacity-planning/-/issues/562#note_1292182723