Issue recovering inverse Hessian from minimize()

[rickecon]

This issue is curated at rickecon/StructEst_W17#26. There are two problems that come up when trying to recover the inverse Hessian matrix from a minimize() function call.

1. Only one of the three constrained minimizer methods ('L-BFGS-B') return the inverse Hessian.
2. Sometimes, the 'L-BFGS-B' method returns the inverse Hessian as a , where K is the number of parameters being estimated.