Leveraging SHAP Values from Machine Learning and Mutational Signatures in Supervised Clustering to Enhance Disease-Specific Survival Prognosis in Endometrial Carcinoma