Towards Transparent AI: A Survey on Explainable Large Language Models