[update] humanoid training pipeline

[myuansun]

Description

Fixes

1. fix the CUDA issue of isaacgym when training walk_g1_dof29;
2. remove redundant physics step in _get_states of isaacgym handler;
3. complete wandb log for rsl_rl_ppo;
4. align the checkpoint loading of roboverse_learn/rl/rsl_rl/eval.py with roboverse_learn/rl/rsl_rl/ppo.py;
5. add ground padding for isaacsim handler to avoid extremely small ground.

Type of change

• Bug fix (non-breaking change which fixes an issue)

How to test

python roboverse_learn/rl/rsl_rl/ppo.py --task walk_g1_dof29 --sim isaacgym --num_envs 2048 --robot g1_dof29 --headless --use_wandb

Checklist

• I have run the pre-commit checks with pre-commit run --color=always --all-files
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have added my name to the CONTRIBUTORS.md or my name already exists there