feat: Added bindings for ProductManifold

[TannerGilbert]

Adds bindings for ProductManifold, needed by the COLMAP IMU demo (colmap/colmap#2625) and more generally useful since colmap::Rigid3d became a parameter block in 4.0.0.

[B1ueber2y]

Thank you!! I m wondering if it is really necessary to rewrite everything instead of directly binding the ceres::ProductManifold out. I guess you tried it but did not manage to get it working. Can you clarify what was the blocker?

On the other hand, re your PR description, this feature is highly useful not only for IMU but also in general as colmap::Rigid3d uses a single parameter block since the 4.0.0 release.

[TannerGilbert]

Thank you!! I m wondering if it is really necessary to rewrite everything instead of directly binding the ceres::ProductManifold out. I guess you tried it but did not manage to get it working. Can you clarify what was the blocker?

On the other hand, re your PR description, this feature is highly useful not only for IMU but also in general as colmap::Rigid3d uses a single parameter block since the 4.0.0 release.

Thanks for the review. Honestly, I'm not super familiar with pybind11, and therefore don't know how to best bind variadic class templates.

template <typename Manifold0, typename Manifold1, typename... ManifoldN>
class ProductManifold final : public Manifold { ... };  

My first try was to just bind ceres::ProductManifold directly, but I couldn't figure out how to make it work and therefore switched to rewriting it. If there is a way to handle class templates like the one above without this I'd be happy to make the changes after having a look.

[TannerGilbert]

Relevant docs: https://pybind11.readthedocs.io/en/stable/advanced/classes.html#binding-classes-with-template-parameters

C++ templates may only be instantiated at compile time, so pybind11 can only wrap instantiated templated classes. You cannot wrap a non-instantiated template

You must explicitly specify each template/type combination that you want to wrap separately.

[B1ueber2y]

I see. Sounds good. Let s go with the current way then. Reminder not to forget to format.

[TannerGilbert]

Tried to implement the changes and also applied formating with clang-format and ruff respectively.

[B1ueber2y]

Thank you. The formatting CI still failed though. Please make sure that your clang-format version matches the one used in this CI (which might be different from what is used in colmap).

[TannerGilbert]

Just checked again and now used the provided .clang-format file.