The Impact of Quantization on Large Reasoning Model Reinforcement Learning