add parquet file output to the data command

[rinaldodev]

We can add some flags later on to:

• Generate a single output file
• Choose the output format (json, parquet, ...)
• Choose compression options
• Optimize things like row group size

This currently outputs one file per component and that might be harder to distribute.

[orpiske]

Thanks! This looks great.

I'm wondering if we should save the files in different directories so we have two datasets (1 in Alpaca Json and the other in parquet) or whether we should keep everything on the same one.

I'm leaning separate, as I am afraid (out of caution) that it could cause trouble setting up some tools.