Join tasks if they are completed

Currently operations are only joined if they are not completed:

Two actions submitted prior to first's completion:

• Client1: Execute Action(action_info)
• Scheduler: Executing(operation_id(1))
• Client2: Execute Action(same_action_info)
• Scheduler: Executing(operation_id(1))
• Scheduler: Completed(operation_id(1))

Action send after the first's completion:

• Client: Execute Action(same_action_info)
• Scheduler: Executing(operation_id(1))
• Scheduler: Completed(operation_id(1))
• Client: Execute Action(same_action_info)
• Scheduler: Executing(operation_id(2))

This should instead be:

• Client: Execute Action(same_action_info)
• Scheduler: Executing(operation_id(1))
• Scheduler: Completed(operation_id(1))
• Client: Execute Action(same_action_info)
• Scheduler: Completed (operation_id(1))

or:

• Client: Execute Action(same_action_info)
• Scheduler: Executing(operation_id(1))
• Scheduler: Completed(operation_id(1))
• Client: Execute Action(same_action_info)
• Scheduler: Cache Check (operation_id(1))
• Scheduler: Completed (operation_id(1))

Start by tracing the scheduler's task-joining path for repeated Execute Action requests, especially how completed operations are handled. Compare the event sequences in the issue and add a regression test showing that a later identical action reuses operation_id(1) or its cached result. Done means the second request no longer creates operation_id(2) and completion is reported correctly.