Generalize predict processes for ML models

[m-mohr]

As discussed in #368, a first draft to generalize the former random forest specific ML processes for predictions into one process for simple class predictions and another process for class probabilities. Please check carefully, I don't have an ML background ;-)

[soxofaan]

Another side note (not necessarily to tackle in this PR): Ultimately, I think, predict_curve becomes obsolete in favor of what is currently named predict_ml_model (considering that fit_curve basically returns just another ML model)

[m-mohr]

Potentially interesting for "bring your own model": https://onnx.ai/

[m-mohr]

We continue in #441 - We still need to bring some of the comments over.