Add reinforcement learning support

[Papaercold]

Summary

• Added a new Reinforcement Learning (RL) module.
• Detailed documentation and usage instructions are available in
  docs/feature/rl_traning.md.

The overall structure and formatting of the RL module follow the same style as the existing state machine data generation module for consistency.

Since end-to-end RL training is inherently challenging and requires careful reward design and tuning, the current implementation focuses on the lift_cube task as a baseline example.

Additional Changes

• Refactored the labels in action_process.py to unify the formatting across:

  • RL module
  • State machine module
  • Teleoperation module

This improves structural consistency and overall code readability.