MILU dataset from AI4Bharat for Indic LLM eval

[abhinand5]

• Raising PR, for the benchmark from this paper [https://arxiv.org/pdf/2411.02538] (https://arxiv.org/pdf/2411.02538)
• This is their git repo -> github.com/AI4Bharat/MILU
• This PR ports the MILU benchmark from AI4Bharat/MILU to the main lm-evaluation-harness repository. The original implementation extends lm-evaluation-harness with YAML task definitions for Indian language evaluation, and consolidating these benchmarks in the main repository would improve accessibility and maintainability for the broader community. This integration aligns with the project's goal of providing comprehensive language model evaluation tools in a centralized location.

[CLAassistant]

All committers have signed the CLA.