Variational Bayes Approximations for Clustering via Mixtures of Normal Inverse Gaussian Distributions