Posterior Approximation by Interpolation for Bayesian Inference in Computationally Expensive Statistical Models