Context-dependent Explainability and Contestability for Trustworthy Medical Artificial Intelligence: Misclassification Identification of Morbidity Recognition Models in Preterm Infants