[tannevaled]

feat(image): say which transform a picture was drawn with

Image carried the box a picture covers on the page, which cannot tell a turned
picture from an upright one: both report the same rectangle. A caller that only
wants to know where a picture is does not care, and one that must map a region
of the PAGE into the picture's own SAMPLES cannot work without it.

That caller is a redaction. ISO 32000-1 12.5.6.23 requires the pixels under a
region to be destroyed rather than masked, so ops has to find them -- and with
the box alone it would destroy the whole intersection of a turned picture's
bounding box, which is more than was asked for. A redaction that removes too
much is its own betrayal.

So Image.Matrix is the transform in force where the picture was drawn, as
a b c d e f, mapping the unit square the picture occupies onto the page. For a
picture inside a form it is the whole chain: the page's transform, the form's
/Matrix and the one at the Do. b and c are zero for an upright picture and stay
zero for a flipped one, so it is also how a caller says "only the straight
ones" and refuses the rest by name rather than guessing.

Three tests: the matrix of an upright, a turned, a skewed and a flipped
placement with the uprightness each implies, and the chain through a form
measured against arithmetic done by hand (5x4 inside a form scaled 3 inside a
page scaled 2 is 30 by 24 at 16, 22).

Co-Authored-By: Claude Opus 5 noreply@anthropic.com