Currently we snap the video's content rect in the local border box space, then snap again in paint container space with paint offset applied. This doesn't make sense and can cause 1px overflow on left and top edge. We should only pre-snap the size, but not location of it.

By the way, we pre-snap the size of video and frame using roundedIntSize. This can cause 1px overflow on right and bottom edge too. Should we switch to flooredIntSize instead?