support building with `BVH_ENABLE_CUDA` enabled

[cz4rs]

Adds BVH_ENABLE_CUDA CMake option and add Cuda workflow to the CI.

• use Kokkos_BitManipulation.hpp instead of relying on intrinsics directly
• use distbvh-cuda11.4.3-gcc9.4-x64 Docker image (it has Kokkos prebuilt with Cuda enabled and Kokkos_ARCH_PASCAL61 selected)
• covers compilation only (as our GitHub runners do not provide GPU)
  • see a helper PR in NGA fork for test output: https://github.com/NexGenAnalytics/distBVH/pull/3
• requires switching to Kokkos::view and other changes described in #11 - follow-up in #16

---

For GitHub GPU runners see: https://github.com/github/roadmap/issues/505 signup.

[nmm0]

Can you rebase? I (finally)) merged the clustering branch