feature request: add support for gpt4all, agieval, and bigbench

[aigeek0x0]

these are most commonly used llm evals and it would be great if you can add these to your benchmark implementations: https://github.com/sgl-project/sglang/tree/main/benchmark