Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution - View it on GitHub