introduce warning for NaN first order matrices

[lombard]

Hi,

this is just a suggestion for future releases.

I spent a couple of hours trying to understand why dynare would freeze with a model that used to solve fine with pre-releases of Dynare 4.

It turned out that having assigned a wrong parameter values was generating NaN values in the jacobian of the model. This was nevertheless passed on to the mex file (mjdgges I guess)... with no return.

Maybe you could enter a "ifnan" and a warning before starting the solution.

Cheers

Gianni