Better Integration of the different ML methods

[QuantumChemist]

As autoplex grew organically with GAP in the beginning, the different ML (hyper)parameters and kwargs etc. are handled very differently. It would be best to refactor the code for making the code more consistent and unify the approaches for the different ML methods.

I would like to suggest that I and @YuanbinLiu work on this.

This issue might take longer to tackle.

TODO:

• take data preprocessing out of ML fit step and make an own step (job?)
• make data flow template (data-gen/build-rss, do_static_dft, collect_dft, preoprocess_data etc.)

[JaGeo]

Agreed. I think for our next steps, the tests seem more important! And this can surely take a bit longer! Most important point for now is that the fitting works 😃