F\`ux\`i: A Benchmark for Evaluating Language Models on Ancient Chinese Text Understanding and Generation