[fixed]OBJ reader and writer disagree on Y texcoords

[hendu]

The OBJ reader flips the Y texcoord by doing (1 - y). The OBJ writer does it by negation, (-y).

This means that processing the same OBJ through irr several times will each time change the Y texcoord: 0 > -1 > -2 > -3...


I believe both approaches are fragile, in properly handling repeating and negative texcoords. For "normal" texcoords that only go 0 to 1, something like this would work: https://github.com/clbr/seirr/commit/1a ... b5ca64d264

[hybrid]

I've only changed the obj writer, as the fabsf would again introduce too many changes in consecutive conversions.