fix(redis): Prevent counter corruption from concurrent mark handled in Redis RQ

[Mantisus]

Description

• Fixes counter corruption on concurrent mark_request_as_handled calls for the same request. When two coroutines concurrently called mark_request_as_handled for the same request, both could pass the hexists check before either executed the pipeline, causing counters to be updated multiple times for the same request, breaking the queue operation.

Issues

• Closes: Negative pending request count when using RedisRequestQueueClient #1873

Testing

• Added a test to verify that counters are updated correctly during concurrent execution.
• Added a test to verify that the request is correctly restored to in_progress after failure.

Checklist

• CI passed

[janbuchar]

And that will result in a retry, correct?

[Mantisus]

Yes, if it's a Redis-related error, it will trigger a retry using @retry_on_error(RedisError).
If something unexpected happens, the error will be propagated.