RLP: Reinforcement as a Pretraining Objective - View it on GitHub