Pretraining & Reinforcement Learning: Sharpening the Axe Before Cutting the Tree