be a little more liberal (tol = sqrt(.Machine$double.eps)) when testing for symmetry of covariance matrices (and make R CMD CHECK monomvn happy again)
