Performance improvements for retessellate

[achirita]

All Submissions:

• [x] Have you followed the guidelines in our Contributing document?
• [x] Have you checked to ensure there aren't other open Pull Requests for the same update/change?
• [x] Does your submission pass tests?

Based on this discussion https://github.com/jscad/OpenJSCAD.org/discussions/1247 I have refactored retessellate.js to improve performance. This function is used by all boolean operations and is the main reason why sometimes they're really slow. To measure the performance I used a geometry with ~15k polygons which gets reduced to ~7.5k polygons. Results for current implementation of retessellate.js :

• total execution time ~580ms of which:
  • ~465ms used to determine coplanar polygons (lines 29-39)
  • ~115ms used to call reTesselateCoplanarPolygons

Results for new implementation:

• total execution time ~120ms of which
  • ~18ms used to determine coplanar polygons (lines 21-22)
  • ~102ms used to call reTesselateCoplanarPolygons

There is an ~5x overall performance improvement, but if we exclude the calls to reTesselateCoplanarPolygons there's more like an ~25x improvement.

I haven't added this test to the existing suite because I didn't want it slowing down build times. If you consider it should be added, I can work something out.

[z3dev]

@platypii wouldn't this be more appropriate in V3?

[platypii]

wouldn't this be more appropriate in V3?

Based on the tests being unchanged, it seems like @achirita implemented this in a way which does not change the output result at all. So it is not a breaking API change.

I think anything without breaking changes should go in V2 still. Why? Because V3 is not ready yet, who knows when it will be ready. But users are actively using V2, so improvements are nice, and usage also let's us find and fix bugs sooner.

[z3dev]

@platypii can you follow with lint changes for V2 modeling?