Optimize ShiftRows() and InvShiftRows()

Optimize ShiftRows() and InvShiftRows()

This switches the code so that it stores the state matrix in the same
column-wise fashion that the FIPS document specifies. This makes it
possible to optimize ShiftRows() and InvShiftRows() by accessing the bytes
inside it as words instead of as bytes. This is guaranteed to work because
the storage allocated by std::vector is guaranteed to be continuous, so
that pointer arithmetic works like normal arrays.

Signed-off-by: Ira W. Snyder <devel@irasnyder.com>