On the Convergence of the Shapley Value in Parametric Bayesian Learning Games