Python engine adapters

[maxwellflitton]

Here I am separating the dependencies out so if a user does just wants to use sklearn then they do not need to have torch and vice versa. I have also isolated the Rust code from the Python code with an adapter pattern so we can now perform unit tests on the processes of training and storing ML files from sklearn and torch and converting them to ONNX. There is still some way to go to get the rust part into unit tests that can be called from Python but for now the rust code itself has unit tests.