[petsc-dev] fix for new MatLoad

[Satish Balay]

On Thu, 8 Jul 2010, Lisandro Dalcin wrote:

> Can any of you replace the PetscValidPointer(newmat,3) check for the
> proper one PetscValidHeaderSpecific(mat,MAT_CLASSID,2); ?

Replaced  PetscValidPointer with PetscValidHeaderSpecific(2).

Also fixed warning with 'prefix'. Previously it was taken from
Viewer. I guess now its from Mat [so make this change]

Satish