check integer overflow in Matrix::setSize

[dreamer2368]

So far, Matrix does not check if the new size is overflowed. This returns a memory error far downstream of application, which takes a long time to backtrace.

Now Matrix::setSize checks if the new_size is overflowed at its execution.

[dylan-copeland]

This check is a good improvement, but I think there is a case in the constructor at Matrix.cpp:133, where setSize is not called, and the check should be added there as well.

[ckendrick]

Thanks for adding this check. This might be a good time to also add this to Vector::setSize as well. Vector has another case where this check should be added to the constructor at Vector.cpp:69.

[dreamer2368]

@dylan-copeland suggestion reflected.

@ckendrick I looked into Vector class, and I think it's actually not the case with the same issue. For Matrix, the size is determined by multiplication inside the class, while for Vector the size is determined simply by a passed input value. We can only check overflow of internal operations. There is no way we can check if the passed dim is overflowed or not: a negative overflown value will be caught by dim > 0; but a positive overflown value, we cannot tell whether that is proper or overflown.

Also for Vector it is easier for users to check if the passed dim is a wrong value, while for Matrix we need to check the internal d_alloc_size.