PoLi-RL: A Point-to-List Reinforcement Learning Framework for Conditional Semantic Textual Similarity