An Empirical Evaluation of the Rashomon Effect in Explainable Machine Learning