pipeline for integrating crowdsource prompts

[sbmaruf]

This PR intends to integrate the following functionality,

• Investigate the stats of the selected dataset. This will help us to estimate the mixing ratio, and translation ratio in the future.
• Download the prompts from the Google sheet.
• jinja template projection (via promptsource).
• Make sure all datasets can be loaded using huggingface load_dataset module.
• Parse g-sheet.
• Automatically validate the prompts.
• Project the prompted dataset. -> Use sbmaruf/project instruct data using psrc #1 instead.
• Automatic data stat generation from google sheet. (suggested by @Kowasaki)

[sbmaruf]

The PR is still in draft.

[Kowasaki]

How are the datasets selected to be included in data_stat.py? Are new additions a manual process? If so I would recommend maybe using something like parsing a .csv file or use the Google Sheets API so that you don't have to manually modify the code everytime a new dataset is added

[sbmaruf]

@Kowasaki That can be done. Let me put it in the TODO.

[AmrMKayid]

docstring please :))

[sbmaruf]

Done. But still WIP.

[sbmaruf]

Estimated finish time: To finish this, some parts of the work need to be finished from the Team-4 collected prompts and review process. But I can still test what we have right now.