Semantics of units employed by [compute containment|cover scaling].

[nigelmegitt]

The requirement for success in the first step of [compute containment scaling] and [compute over scaling] is that the units of TR and RR are the same, but this is rarely likely to be the case.

Propose replacing with a step to re-express image dimensions in the document coordinate space so that the resulting units are always the same.

In terms of units, there's no way any image resource will sensibly have, say, rw or % units, unless some processing step upstream has deliberately taken the intrinsic dimensions and re-expressed them in those units to get to TR. If the intent is to constrain document instances to express positions and sizes in pixels if they use embedded images, it's a strange way to achieve that goal.

Originally posted by @nigelmegitt in #1098