Fix various operators for padding when using SIMD/AVX
This MR fix padding problems with the following operators:
- AssmbScatrStdMat
- NullPreconStdMat
- DiagPreconStdMat
- DirBndCondStdMat
- RobBndCondStdMat
- NeuBndCondStdMat
- ConjGradStdMat
- FwdTransStdMat
- HelmSolveStdMat
Note: The FwdTrans test is failing due to a tolerance issue when using AVX. To be fixed.
Edited by Jacques Xing