Facets of regularization in high-dimensional learning: Cross-validation, risk monotonization, and model complexity