The old implementation did a standard bilinear/"quad" warp on... the
Bezier control points.
That's not correct. A much better approximation of warping the points on
the curve itself can be achieved by first subdividing the curve into
smaller segments over which the bilinear warp is closer to affine. So,
the error is lower. Why did I do this? Because the grid pattern
indicates it's a bilinear warp and I was very confused when it wasn't.
Change-Id: I9478007c8662455ddbb56fa9d7e6dff425c8e718
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/210921
Reviewed-by: Regina Henschel <rb.henschel@t-online.de>
Tested-by: Jenkins
Reviewed-by: Ilmari Lauhakangas <ilmari.lauhakangas@libreoffice.org>
