improve _zoomSurfaceY speed
Brought to you by:
ferzkopp
i find out that replacing `sp += (*csax) * (flipx ? -1 : 1);` from line 589 by `sp += (*csax)` and adding new line `csax*= (flipx ? -1 : 1);` after 559, improve greatly speed of that function.
i think that `_zoomSurfaceRGBA` have similar problem that can be solved by removing redundant operation (that dont depend on row position in graphic) from inner loop.
Thanks! Fixed for _zoomSurfaceY, but the _zoomSurfaceRGBA algorithm has changed and does not need to be fixed anymore like this.