Draft: fix: unset SidekiqStatus on job failure

Changes

Fixed lib/gitlab/sidekiq_status/server_middleware.rb

• Replaced the current implementation with a begin-ensure block pattern
• The ensure block guarantees that SidekiqStatus.unset(job['jid']) is always called, even when exceptions are raised
• This follows the same pattern used in other middleware like BatchLoader
• The implementation is now cleaner and more concise (4 lines instead of 6)

Added test coverage in spec/lib/gitlab/sidekiq_status/server_middleware_spec.rb

• Added a new test case: "stops tracking of a job even when an error is raised during job execution"
• The test verifies that SidekiqStatus.unset is called even when the job raises an exception
• Follows the same test pattern used in extra_done_log_metadata_spec.rb

Problem Solved

• Before: If a job raised an exception, unset was never called, leaving orphaned status keys in Redis for 30 minutes
• After: The ensure block guarantees cleanup happens regardless of whether the job succeeds or fails
• This prevents SidekiqStatus.running?(job_id) from incorrectly returning true for failed jobs