Added new HF datasets => Perplexity format functions

[tusharkhot]

New datasets: arc-easy, boolqa, obqa, hellaswag, mmlu, xsum, imdb.

Uploaded converted files to: s3://ai2-llm/eval-data/downstream/v0/

@IanMagnusson for visibility. Feel free to merge whenever convenient.

[tusharkhot]

@IanMagnusson Added the remaining downstream evals. Also uploaded the generated files to S3