ENH introduce SKRUB_DATA_DIRECTORY envar to control the data directory

[thomass-dev]

Closes #1216.

Adding a new way to control the location of the data directory, using envar.

It can be useful when you want to set the data cache directory from outside the code, especially in CI.
In addition, fix the expansion of ~ when passing programmatically an explicit path to get_data_home.

[thomass-dev]

Can i instantiate a logger from stdlib?
I've seen the use of loguru in the benchmark folder, so i reproduced, but i see that it's not a dependency.

[glemaitre]

loguru will not be a dependency for the main package (the benchmark lives on its own).

I see that we are using logger:

https://github.com/probabl-ai/skore/blob/418b72ca6711ef467c95f67416f2162382c645a7/skore/src/skore/__init__.py#L21-L23

I did not recall because in scikit-learn, we using some verbose parameter + print function. But since, the logger is here, it should be used I assume.

@jeromedockes can probably confirm it

[glemaitre]

Whoops, it is actually late and I'm confusing the sk*** pacakge. So we don't use logger which is more what I thought. I would expect that we use verbose + print.

@jeromedockes could confirm

[glemaitre]

we will need a small test to check that we write in the good directory as well. I think that we only need to test the get_data_home for that.

[jeromedockes]

thanks!

[jeromedockes]

the tests failure are not related to your PR but to #1217 :/

[thomass-dev]

@jeromedockes i have fixed your comments.
In addition, i have removed the print statement. It makes no sense to print the data_home each time the function is called.

[thomass-dev]

The PR is up-to-date and ready for review.

[jeromedockes]

thanks a lot @thomass-dev ! sorry you needed to update it after fixing a bunch of conflicts and thanks for addressing all comments

[jeromedockes]

oops sorry I forgot one last detail: if you could add a brief entry with your name and github handle in the changelog that would be great:

skrub/CHANGES.rst

Line 12 in 1831731

New features

thanks!

[thomass-dev]

Thanks @jeromedockes , it should be okay?

[jeromedockes]

yes, perfect. thanks!