Benchmarking Large Language Models for Persian: A Preliminary Study Focusing on ChatGPT