ENH: Add download_all utility method & script

[AnirudhDagar]

Reference issue

#16983

What does this implement/fix?

_download_all.py can be run as a standalone script to download all the data files required by scipy.datasets at once without the need to build SciPy first. This could be useful; for example, SciPy Debian, Fedora etc. packages may have build time constraints for downloading from external links.

As per the discussion with Ralf earlier today, the steps might look something like this:

git clone ...
git submodule update ...
# Download dataset files
python _download_all.py 

...continue the build process

Although I couldn't find the exact SciPy Debian build script in https://salsa.debian.org/python-team/packages/scipy, where something like this should be placed, it would be great to hear @drew-parsons, your thoughts and if could help us review this. Thanks!

Besides, download_all can also be used as a utility method within the scipy.datasets module if required.

from scipy import datasets

# Cache all datasets
datasets.download_all()

# Now work without an internet connection

cc @rgommers

Additional Information

Sharing Network access by Debian package builds by Nathan Willis, talks about some constraints when using the internet at build time in Debian packages for people who didn't know about this strict rule.
I found it a nice read since I recently learnt about the constraint when working on scipy.datasets module.

[AnirudhDagar]

@rgommers, a little reminder requesting your review, in case it was missed. I'll raise the clear_cache utility method PR (already implemented) after this lands to avoid unnecessary merge conflicts. Thanks!

[rgommers]

Overall LGTM, a couple of changes needed @AnirudhDagar. And apologies for the delay in reviewing

[rgommers]

LGTM now, thanks @AnirudhDagar. In it goes.