Adjust wrong reactive cache workers retry amount

What does this MR do?

Context

The fact that models/concerns are loaded before initializers (for production) led to a bug where an early evaluation of these modules (and cascading to evaluation of ReactiveCachingWorker and ExternalServiceReactiveCachingWorker) would use an outdated Sidekiq.sidekiq_options (which is set to retry: true by default on Sidekiq).

This allow these workers in production to retry up to 25 times, spending quite some time doing it.

In summary, the bug consisted in:

1. These worker classes were initialized with retry: true Sidekiq configuration, given the initializers/sidekiq.rb wasn't evaluated yet
2. The initializer set the default properly
3. The subsequently loaded workers would get the right default

gitlab-com/gl-infra/scalability#356

Screenshots

Does this MR meet the acceptance criteria?

Conformity

• Changelog entry
• Documentation (if required)
• Code review guidelines
• Merge request performance guidelines
• Style guides
• Database guides
• Separation of EE specific content