Approaches to the Interpretation of Machine Learning Models Trained with Big Experimental Kinetic Data: An Example of the Suzuki–Miyaura Reaction