Benchmarking Large Multimodal Models for Ophthalmic Visual Question Answering with OphthalWeChat