When predictions fail, especially within CF's use of Optaplanner TrustyAI gives a very useless "Prediction failed" message. If we catch a Model predictAsync error within Python and pass it to a log before it reports back to the JVM we don't lose the original Python error message