[Refactor] Clear `run_grpo_math.py` and `run_grpo_rm.py`

[RayenTian]

Description
This PR #1506 introduces a new executable script: run_grpo.py.

The main purpose is to consolidate and generalize functionality that is currently split across run_grpo_math.py and run_grpo_rm.py.

After #1506 successful integration and confirmation of completeness, the obsolete scripts will be removed in a separate PR.

Related PRs
#1506