Add documentation for PGLearn

[klamike]

The docs should have some tutorial for how to download and use PGLearn in ML4OPF.

[klamike]

HF has removed support for custom dataset scripts: huggingface/datasets#7592 (comment)

cc @mtanneau

[mtanneau]

How does that affect us? No more ability to auto-generate croissant files?

[klamike]

We can still do it using an old version of datasets. But yes, with latest datasets, the upload pipeline will not work since it depends on convert_to_parquet which is gone.

Since we converted to parquet before that change, users can still do load_dataset like normal though (I think, I haven't tested it). The manual download (which we recommend) should still work fine.

I've started huggingface/datasets#7690...