Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Models - View it on GitHub