Fix issue in calculating NDCG when input data is smaller than K

[oriol-guitart-edo]

This PR tries to fix the open issue which appears when some users have less recommendations than the K

[emeli-dral]

Hi @oriol-guitart-edo ,
Thank you for your contribution!

I believe reasons of failed tests are formatting and a missing bracket.
I suggest you try to fix the 55 line in the file like that: idcg = df.groupby("users").target.apply(lambda x: x[:i].dot(discount[: len(x)])).rename("idcg") (add and extra bracket after "len(x)]" and add a space in "[: len(x)]")

[oriol-guitart-edo]

Yes, it is fixed. Hopefully, it will pass the tests now

[emeli-dral]

Looks like there is still an issue with black formatter.
Here is a way how you can check it locally: black -l 120 -t py38 ndcg_k.py
I believe it is only a missing whitespace here: [: len(x)]