Maintain manifoldness when converting between PolySet and ManifoldGeometry

[kintel]

Fixes #5022

• createManifoldFromPolySet() is now the main entry point for obtaining a ManifoldGeometry from a PolySet
• We first try direct creation from vertices and indices (after triangulating if necessary)
• Added ResultObject::mutableResult() and ResultObject::constResult() to make it more clear what to expect, plus clean up some wrong constness.
• Fixed a test case with wrong polygon winding: tests/data/scad/stl/stl-export.scad

[kintel]

@pca006132 This is not mergeable, but could be a starting point for experimentation/discussion.

[pca006132]

Wow this is a pretty large patch, I think we can probably start small and try to construct ManifoldGeometry from some primitives and polyhedron calls?

[kintel]

Oh, don't worry, just look at the very last commit.

[pca006132]

Seems fine, not sure how the status thing will interact with multiple threads though (not that it matters now). I think manifold should probably provide some thread safe alternative to check the status, though it may be a hard problem (to keep compatibility).

[kintel]

What part of the status handling is not thread safe?

[pca006132]

Ah, nevermind, I remembered the wrong API. I thought that status is static but it is not

[kintel]

@pca006132 If you find the time: The interesting pieces are createManifoldFromPolySet() and createManifoldFromTriangularPolySet()