Learning to Use Tools: Reinforcement Learning for Tool-Integrated Mathematical Reasoning