Truncated Multivariate Normal Generator

(perfect) sampling from the truncated multivariate normal density

Simulates 'n' random vectors exactly (perfectly) distributed from the d-dimensional N(0,Sig) distribution (zero-mean normal with covariance 'Sig'), conditional on l<X<u. Infinite values for the truncation limits 'l' and 'u' are accepted.
Reference: Z. I. Botev (2015), "The Normal Law Under Linear Restrictions: Simulation and Estimation via Minimax Tilting", submitted to JRSS(B)