in branch mlflow-new-model-testing, load the model path from config

[jsvobo]

Add config options for model and scaler path. can train different models at once in slips (run different instances with their config)