fix: update python decompose example

[AngeloDanducci]

• The python decompose example needs a larger model to guarantee tags are created appropriately for subtasks.
• Ref update to the cli needs to be updated.
• Changes to the decompose pipeline in fix(mellea decomp): Solve ConstraintExtractor parsing fails and improve robustness #445 have made it so that subtask_data.constraints may now contain constraint strings that were not in the original task_prompt_constraints list (they're sub-parts of split constraints). Initial approach to get the test working again will be to simply assign llm as the validation strategy if none exists. Though perhaps elsewhere in the pipeline it should instead map parent -> child (split constraints) validation strategy.

[psschwei]

FWIW, I tried running with granite3.3:8b this morning and got this error:

Traceback (most recent call last):
  File "/Users/paulschw/generative-computing/mellea/docs/examples/m_decompose/python/python_decompose_example.py", line 229, in <module>
    main()
  File "/Users/paulschw/generative-computing/mellea/docs/examples/m_decompose/python/python_decompose_example.py", line 208, in main
    result = run_decompose(task_prompt)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/paulschw/generative-computing/mellea/docs/examples/m_decompose/python/python_decompose_example.py", line 33, in run_decompose
    result = decompose(
             ^^^^^^^^^^
  File "/Users/paulschw/generative-computing/mellea/cli/decompose/pipeline.py", line 119, in decompose
    cons_key: validation_decision.generate(m_session, cons_key).parse()
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/paulschw/generative-computing/mellea/cli/decompose/prompt_modules/_prompt_modules.py", line 34, in parse
    return self._parser(self.__str__())
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/paulschw/generative-computing/mellea/cli/decompose/prompt_modules/validation_decision/_validation_decision.py", line 70, in _default_parser
    return _ValidationDecision._assert_output_format(normalized_decision)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/paulschw/generative-computing/mellea/cli/decompose/prompt_modules/validation_decision/_validation_decision.py", line 31, in _assert_output_format
    raise AssertionError(
AssertionError: LLM generated invalid output: "llm
all tags are closed and my assignment is finished.". Expected either "code" or "llm".

[AngeloDanducci]

See. my comment in the mentioned PR, this fixes that issue as well.

[planetf1]

Addressed by PR #569 (fix: python decompose model change and pipeline fix).

[AngeloDanducci]

Didn't specify "fixes" in the pr oops, correct we can close this one.