Differentiability and Approximation of Probability Functions under Gaussian Mixture Models: A Bayesian Approach