Jacobian for Mean/StdDev GMM estimation
[M,e] = gmmMsdJ(b,gmmopt,Y,X,Z)
* b = parameter vector fed to function
* gmmopt = gmm options tlist
* Y = "dependent" variables
* X = "independent" variables
* Z = instruments (can be same as X)
* M = Jacobian (k+k(k+1)/2 by k+k(k+1)/2) (k is cols(y)
* e = model errors (Nobs x Neq)