Clipper2 Julia wrapper

[gpeairs]

This is the companion to #60 adding the Julia side of the wrapper -- it would make sense to me to review/merge that one separately, but this is close to complete from my perspective and should help with evaluating the C wrapper.

The libcclipper2 product is being added to Clipper2_jll by JuliaPackaging/Yggdrasil#14270, which is gated on #60. Until that JLL is published, place a matching local library at deps/libcclipper2.so, deps/libcclipper2.dylib, or deps/libcclipper2.dll for local testing. It must be built from deps/cwrapper/cclipper2.cpp with -DUSINGZ against the patched Clipper2 2.0.1 used by that Yggdrasil recipe.

The package loader and Project.toml will use the Clipper2_jll.libcclipper2 product once it is available. CI will fail until then.

I've marked this as a draft for now, since it needs to wait until the above PRs are merged.

[SimonDanisch]

Let an LLM try this pr out: built the wrapper against patched Clipper2 2.0.1 per Yggdrasil #14270 — 112/112 tests pass, docs clean. Fuzzed ~1000 boolean ops and stress-tested GC in the result callbacks; no issues.

One defect: Clipper64() / ClipperOffset() don't check the NULL return documented at cclipper2.h:79 (engine.jl:71, offset.jl:23), so a failed create segfaults on first use instead of throwing.

One question: the *_z entry points have no Julia surface: only runtests.jl reaches them via raw ccall. Should those be wrapped?

[gpeairs]

Thanks for trying it out!

One defect: Clipper64() / ClipperOffset() don't check the NULL return documented at cclipper2.h:79 (engine.jl:71, offset.jl:23), so a failed create segfaults on first use instead of throwing.

Right, fixed.

One question: the *_z entry points have no Julia surface: only runtests.jl reaches them via raw ccall. Should those be wrapped?

Yeah, they should. Added Z versions of the types, dispatch on Z paths for the add_...! methods, and execute_polytree_z.