Add link to PharmKG8k about alternate version of data

[skingi20]

The PharmKG8k dataset that you use, and the one which the people behind the PharmKG8k paper point to have some significant issues.

We think that we have solved these significant issues. The issues consist of data leakage and the split being inductive.
More can be read about it on this GitHub:
https://github.com/skingi20/improvement_of_PharmKG8k_split
The new split that we propose, which has no data leakage and is transductive, can be found on there as well as proof of the original PharmKG8k split's problems.

[cthoyt]

I'm a domain expert in biology and chemistry. I'm not very happy with any of the "benchmark" biomedical knowledge graphs we have added in PyKEEN. Despite that, they're included for the convenience of people who aren't so skilled with data acquisition.

I would suggest not investing any time in PharmKG nor its variants since it is one of the worse ones, both in terms of the content included, the questionable semantics, and the fact that its corresponding paper doesn't explain much about how it's made (plus no code available).

If you want to send a PR that adds a warning in the docs for PharmKG that there are issues and people should consider your alternative, then we could consider that, but making more variants of a questionable resource probably won't help pykeen users in the long run.