ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning