validate env scoring mechanism properly ie rubrics actually give different score

[examples/qed_math_inference.py](https://github.com/rycerzes/OpenEnv/blob/389980d73452a892dba6a9af52faab09c2f0d285/examples/qed_math_inference.py)