Update rsl-rl-lib version constraint to fix ValueError when start training.

[CMBill]

开始训练时报错：

Traceback (most recent call last):
  File "/home/c112/codes/LeggedLab/legged_lab/scripts/train.py", line 101, in <module>
    train()
  File "/home/c112/codes/LeggedLab/legged_lab/scripts/train.py", line 85, in train
    runner = OnPolicyRunner(env, agent_cfg.to_dict(), log_dir=log_dir, device=agent_cfg.device)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/c112/miniconda3/envs/LeggedLab/lib/python3.11/site-packages/rsl_rl/runners/on_policy_runner.py", line 45, in __init__
    self.cfg["obs_groups"] = resolve_obs_groups(obs, self.cfg["obs_groups"], default_sets)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/c112/miniconda3/envs/LeggedLab/lib/python3.11/site-packages/rsl_rl/utils/utils.py", line 252, in resolve_obs_groups
    raise ValueError(
ValueError: The observation configuration dictionary 'obs_groups' must contain the 'policy' key. Found keys: []

后尝试使用低于3.0版本的rsl_rl包测试发现没有报错，应该是rsl_rl包的版本问题，于是修改了setup.py中install_requires里对rsl-rl-lib的版本要求，添加了<3.0.0。

目前不确定较早的3.0版本的rsl_rl包是否报错，但是2.0版本中最新的2.3.3是没问题的。

[Copilot]

Pull request overview

This PR fixes a ValueError that occurs when starting training with rsl-rl-lib version 3.0.0 or higher by adding an upper bound version constraint to exclude incompatible versions. The error was caused by breaking API changes in rsl-rl 3.0+ that expect an 'obs_groups' configuration with a 'policy' key that is not being provided by the current codebase.

Key Changes:

• Constrained rsl-rl-lib dependency to versions >=2.3.0 and <3.0.0 to avoid incompatibility issues with version 3.0+

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CMBill]

found issue #38