remove or replace `cgeo` C++ module

[abc8747]

minisky/tools/src_cpp/setup.py uses distutils.core.Extension, which was deprecated in PEP 632 (Python 3.10). Since we have to build cross-platform wheels soon anyway, we should look into either:

• use scikit-build-core, specifically pybind11/nanobind (cmake is a headache though)
• or rewrite it in Rust altogether, and use PyO3/maturin so we can generate .pyi stubs easily. However, compatibility with numpy arrays is unknown (maybe we can borrow some ideas from @xoolive 's thrust)

The latter approach is desired because we the current tangram frontend lacks any syntax highlighting for commands, so maybe we can create a wasm module with Rust. But whether that is really needed is still too early to tell

[amorfinv]

Hello @abc8747, in regular BlueSky I rarely used the cpp tools. So there could be an argument to just remove minisky/tools/src_cpp

There would be less things to maintain and the code is quite old.

[abc8747]

@amorfinv Oh sure, we can do that too!

I guess we should profile the code to understand the hotspots and see if switching on the C++ geo module brings any material speedup. If there isn't any major speedup, we should just remove it to avoid the headache of building wheels for all platforms.

Quick note: I removed the prefer_compiled option in the settings.toml in #28, it should just try to import the compiled extensions (which would fail anyway because distutils is broken) otherwise fall back to Python.

[amorfinv]

@abc8747 Yeah profiling is a good idea. From experience conflict detection and resolution is one large bottleneck since it was difficult to vectorise.

[xoolive]

If that's the part of code I am thinking about, and esp. if there's other compiled parts in Rust, I would suggest to reuse our existing pieces of code for geodesy here, or use new libraries. The landscape of geodesy libraries has changed a lot since the first versions of Bluesky

[amorfinv]

Then I suggest we remove the cpp tools before first release and work on implementing rust geodesy in future