Double precision rational approximation algorithm for the inverse standard normal first order loss function