Inherently interpretable machine learning for credit scoring: Optimal classification tree with hyperplane splits