Rename The Pile subsets

[lhoestq]

After discussing with @yjernite we think it's better to have the subsets of The Pile explicitly have "the_pile" in their names.

I'm doing the changes for the subsets that @richarddwang added:

• books3 -> the_pile_books3 add books3 #2801
• stack_exchange -> the_pile_stack_exchange add stack exchange #2803
• openwebtext2 -> the_pile_openwebtext2 add openwebtext2 #2802

For consistency we should also rename bookcorpusopen to the_pile_bookcorpus IMO, but let me know what you think.
(we can just add a deprecation message to bookcorpusopen for now and add the_pile_bookcorpus)

[thomwolf]

Sounds good. Should we also have a “the_pile” dataset with the subsets as configuration?

[lhoestq]

I think the main the_pile datasets will be the one that is the mix of all the subsets: https://the-eye.eu/public/AI/pile/

We can also add configurations for each subset, and even allow users to specify the subsets they want:

from datasets import load_dataset

load_dataset("the_pile", subsets=["openwebtext2", "books3", "hn"])

We're alrady doing something similar for mC4, where users can specify the list of languages they want to load.