On Quantifying Literals in Boolean Logic and Its Applications to Explainable AI