Towards Understanding Fairness and its Composition in Ensemble Machine Learning