Positive-Only Drifting Policy Optimization

Create an easy-to-use RL library that implements the PODPO algorithm for training policies without gradient clipping. It would simplify the training pipeline for complex control systems.

Suggested repo: nanoRL

"Reinforcement learning without the gradient clipping headaches."

Estimated effort: 60h