Sort holes in interiorcuts

[gpeairs]

Fixes #175.

Right now _normalize_clip_arg turns ClippedPolygons into polygons via interiorcuts, but we can just extract the oriented contours directly. This then lets us do the guillotine cut for ClippedPolygons with more points than GDS_POLYGON_MAX without an intermediate interiorcuts, which fixes the mentioned issue. [Edit: it doesn't, see comments further down on how we actually fix the issue.]

In a follow-up, we can probably do the same for offsets of ClippedPolygons, which should address the issue where shrinking a polygon with an interior cut pulls it open. [This is still probably a good idea.]

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[gpeairs]

We need a solution that works for rendering styled ClippedPolygons, but I don't want to start pulling apart the to_polygons -> render! separation. Maybe it's enough to self-union any big polygon so it takes the ClippedPolygon guillotine path anyway?

[gpeairs]

Looks like the guillotine cut interaction was a red herring -- interiorcuts was just processing holes in a bad order. I'm surprised we hadn't encountered this before, but sorting contours by the y coordinate of the cut start point should guarantee that no cut passes through a hole that hasn't been processed into the outer contour yet.

[laylagi]

LGTM, it fixes the issue in my example. Thank you!