Cardinality-Minimal Explanations for Monotonic Neural Networks