Add MBPP

[hjlee1371]

Hi, I added the widely-used MBPP benchmark. This partially resolves #1157.

Similar to #1992 , the implementation relies on pass@k from the HF evaluate module, so it requires the environment variable HF_ALLOW_CODE_EVAL=1.

Below are results for some important pretrained models, along with scores reported from the llama3 and gemma2 papers. Also note that the prompting follows the original paper, which is different from the bigcode-eval.

Models	3-shot MBPP pass@1 (lm-eval)	reported from llama3	reported from gemma2
Meta-Llama-3-8B	46.0	-	-
Meta-Llama-3.1-8B	47.0	47.6	-
gemma-7b	44.8	44.4	44.4
Mistral-7b-v0.1	37.8	47.5	40.2

[go2ready]

Hello! Any blockers for adding MBPP?

[baberabb]

thanks for the PR!