The Multivariate Saddlepoint Approximation to the Distribution of Estimators: A General Approach