Make BGRA deinterlacing benchmark premultiplied, so that we do not use forever just...

[movit] / multiply_effect.frag

vec4 FUNCNAME(vec2 tc) {
        return INPUT(tc) * PREFIX(factor);
}