Reinforcement learning to enable reasoning LLMs for Text2SQL