Towards Comprehensive Detection of Chinese Harmful Memes