✨ feat: 新增 MJX/brax-PPO 训练管线 (lab 6 pupper)

- envs/pupper_env_mjx.py — MJX baseline 环境（36-dim obs × 15 frames,
  18 reward terms, 域随机化, cmd vx±0.75/vy±0.5/wz±2.0）
- train_brax_ppo.py — brax-PPO 训练入口，tensorboardX + jsonl 双写指标
- envs/__init__.py — guard import，允许 brax-only 环境中加载

验证：200M 步训练（num_envs=8192, batch_size=256），约 149M 步处达到
峰值 ep_rew 51.15 ± 7.4，存活率 100%。

Author: bokang-ugent

codes/practices/quadruped/cs123/exercises/lab_6_rl_pupper/envs/__init__.py

@@ -1,3 +1,11 @@
-from envs.pupper_env import PupperEnv
+"""Lab 6 RL Pupper env exports.
 
-__all__ = ["PupperEnv"]
+Eager `PupperEnv` import is guarded so this package can be imported in
+MJX/brax-only environments where gymnasium / stable_baselines3 are absent.
+"""
+
+try:
+    from envs.pupper_env import PupperEnv  # noqa: F401
+    __all__ = ["PupperEnv"]
+except ImportError:
+    __all__ = []