Developing UMAC: A Unified Model-Agnostic Computation Process for Enhanced Machine Learning Explainability