glifo: Simplify COLR glyph handling

[LaurenzV]

Initially, in Vello CPU, any COLR glyph would first be rendered into an intermediate pixmap and that pixmap would then be rendered into the surface. Because of this, in case there was a skewing factor in the glyph, instead of just rasterizing the glyph at its normal font size, we would apply an additional scaling factor such that it would be rasterized at a higher solution, avoiding aliasing artifacts that arise from the skewing.

However, since the most recent changes, things are different: By default, glyphs are always directly rendered into a surface. Glyphs are only cached if they don't have any skewing factor (if in the future we decide that we also want glyph caching for rotated glyphs, we can easily just fold this into the font size again). Therefore, this additional scaling procedure becomes unnecessary. By doing so, the code becomes a easier to understand (in my opinion).

As a nice side effect, this PR also fixes a discrepancy between cached and non-cached rendering that existed previously:

rec.mp4

As you can see, previously (on the left) cached and non-cached glyphs would have slightly different positions. Now, they seem to have the same position. I must confess I don't fully understand why this was the case, but the tests don't lie! 😄

[nicoburns]

Does this apply to regular skewed glyphs? I imagine skewed COLR glyphs are pretty uncommon. But skewed regular glyphs for "fake italic" would be significantly more common.

[LaurenzV]

Nope, it only applies for cached glyphs. And caching i never applied if a glyph is skewed.

[LaurenzV]

Sorry, that was an accident 😅