Create pheval.experimental runner for cross species testing

[matentzn]

Note, this is not the same as #173, where we are looking into creating a simple default runner for testing the framework that does basically nothing.

As described in #206 our goal is to figure out the WNATO hypothesis. We will use pheval.experimental as a nice experimental runner for any number of random shenanigans we want to experiment with (note there is some overlap with #159 for LLM related stuff which is worked on by @yaseminbridges, but lets not worry about scopes here).

I would suggest the following:

1. @kcortes133 you implement a pheval runner under your own name, called pheval.experimental (like https://github.com/monarch-initiative/pheval.exomiser), following the instructions in https://monarch-initiative.github.io/pheval/ @yaseminbridges and @souzadevinicius will certainly help (@yaseminbridges you took some notes during your last implementation right?)
2. As a first pass, we should have a method that essentially wraps grape link prediction. (@kcortes133)
3. I would like a simple method that just works of semantic similarity (no ML at all), which essentially matches phenotypic profiles to known disease profiles (nothing else). The semantic similarity profiles are passed in rather than computed on the fly (@souzadevinicius)