Improve testing for context serialization

[yannlambret]

Follow-up to #50446 and #50521:

• Clean up requirements-flattening logic.
• Add tests for requirements passed as multi-line strings.

Following @shahar1 PR (#50521), I cleaned up a little bit the flattening logic of
_iter_serializable_context_keys, in connection with this part of the __init__ function
of PythonVirtualenvOperator class:

        if not requirements:
            self.requirements: list[str] = []
        elif isinstance(requirements, str):
            self.requirements = [requirements]

As self.requirements is always a list starting from there, it simplifies the way we can process it later. I also add the use cases detailed in #50521 to the relevant test. Sorry for my initial over-simplification 🙏🏻

[amoghrajesh]

No problem! Thanks for taking this on.

[amoghrajesh]

LGTM

[amoghrajesh]

Test failures are unrelated:

  providers/atlassian/jira/tests/unit/atlassian/jira/operators/test_jira.py s [ 10%]
  sss                                                                      [ 10%]
  providers/atlassian/jira/tests/unit/atlassian/jira/sensors/test_jira.py s [ 10%]
                                                                           [ 10%]
  providers/celery/tests/unit/celery/cli/test_celery_command.py .          [ 10%]
  providers/celery/tests/unit/celery/executors/test_celery_executor.py ... [ 10%]
  .sss.
  Test failed with 143. Dumping logs

rebasing it

[yannlambret]

Hi @amoghrajesh, I've rebased my branch cleanly onto main (dropped the merge commit and cherry-picked my changes). Hope the tests pass 🤞🏻

[yannlambret]

All good!

[yannlambret]

Hello @eladkal, could you please review this?