Narrow AirflowException to specific exceptions in DataSync operator

[haseebmalik18]

The AWS DataSync operator raised the catch-all AirflowException for ten unrelated conditions, so a DAG author could not distinguish an invalid constructor call from a failed task execution when handling errors.

This narrows each usage to a specific type:

• Usage/validation errors now raise ValueError.
• Runtime domain failures raise dedicated exceptions added to
  aws/exceptions.py:
  • DataSyncTaskNotFoundError
  • DataSyncMultipleTasksError / DataSyncMultipleLocationsError
  • DataSyncLocationNotFoundError
  • DataSyncTaskCreationError
  • DataSyncTaskExecutionFailedError

The new classes subclass AirflowException, so existing except AirflowException handlers keep working, with no breaking change.

Tests are updated to assert the specific types, with added coverage for the raise sites that were previously untested.

Part of the ongoing effort to reduce direct AirflowException usage. The check-no-new-airflow-exceptions allowlist entry for this file is tightened from 10 to 0.

[Sanjays2402]

switching this from AirflowException to ValueError is a behavior change for anyone catching AirflowException around operator construction, since ValueError isn't in that hierarchy. the raises in execute() all stay AirflowException subclasses, so this one's now the odd one out. intentional?

[haseebmalik18]

Yeah, intentional. The constructor check is argument validation, so ValueError fits it better, while the execute failures stay AirflowException subclasses. It's also at parse time where the DAG loader catches broad Exception, so a bad config still surfaces as a normal import error.