Similar Items: Towards Generalisable and Explainable Traffic Signal Control via Deep Reinforcement Learning and Large Language Models