Add template for parquet datasets

[klei22]

Adding scripts compatible with:

1. Canola - english-python dataset from webcrawled data
2. MMLU-PRO - a benchmark of difficult multple choice questions

Also created

get_json_dataset.py - for getting json datasets automatically from a url get_parquet_dataset.py - for getting parquet datasets automatically from a url

And a helper script: get_dataset.sh from the template folder with prefilled fields for easily getting new datasets.