Module aa 

AA pipe: shape byte present, BlendMode::Normal, no soft mask, isolated group.

Equivalent to Splash::pipeRunAA{Mono8,RGB8,XBGR8,BGR8,CMYK8,DeviceN8}.

For each pixel:

1. a_src = div255(a_input * shape) — scale source alpha by AA coverage.
2. If a_src == 255: direct write (no read-back needed).
3. If a_src == 0 and a_dst == 0: write zeros.
4. Otherwise: a_result = a_src + a_dst - div255(a_src * a_dst). c_result = ((a_result - a_src) * c_dst + a_src * c_src) / a_result. Then apply transfer LUT.