Decision Information Meets Large Language Models: The Future of Explainable Operations Research