Seeing It All: Evaluating Supervised Machine Learning Methods for the Classification of Diverse Otariid Behaviours