Back to the Future: Towards Explainable Temporal Reasoning with Large Language Models