Fix swallowed CancelledError in start_child_workflow and Nexus operations (Issue #1445)

[yegorske50]

What was changed

Added the raise statement to the asyncio.CancelledError except blocks within _outbound_start_child_workflow and _outbound_start_nexus_operation in temporalio/worker/_workflow_instance.py.

Why?

Currently, if a cancellation request occurs while a parent workflow is yielding to start a child workflow or Nexus operation, the asyncio.CancelledError is caught to apply the cancel command, but it is not re-raised. Because this happens inside a while True loop, the error is swallowed, causing the parent workflow to completely ignore the cancellation and hang or proceed incorrectly. Re-raising enforces standard asyncio cancellation propagation and allows the workflow to shut down gracefully.

Checklist

1. Closes [Bug] CancelledError is swallowed during workflow.start_child_workflow #1445

2. How was this tested:

• Created a local reproduction script using an empty task queue to force a hang on start_child_workflow, verifying the cancellation was previously ignored.
• Verified that adding the raise statement allows the parent workflow to successfully catch the CancelledError and shut down.
• Ran the local test suite (python -m pytest tests/worker/test_workflow.py), confirming no regressions in core workflow execution or deadlock detection.

3. Any docs updates needed?
   None.

[CLAassistant]

All committers have signed the CLA.

[tconley1428]

I don't believe this fix is correct, as asyncio.CancelledError is not guaranteed to come from cancellation of the worker. You should check that cancellation of the worker has actually been requested.