Implement cuda.parallel Python module

[wmaxey]

We've just merged cuda.cooperative Python module. Next step is prototyping cuda.parallel module with host-side (device-level) parallel algorithms. Prototyping is scoped to device reduction. Remaining algorithms are about to be tracked by standalone issues.

Tasks

Preview Give feedback

1. Extract device reduction kernels into NVRTC-compilable header #2230
   +0
   
2. Implement C API for device reduction #2232
   +0
   
3. [FEA]: Move cuda cooperative module #2268
   feature request +1
   
4. FEA: Introduce CMake / CI infrastructure for cccl/c #2235
   3 of 3
   +0
   
5. Introduce Python binding to device reduction #2236
   +0
   
6. Introduce CI infrastructure for cuda.parallel #2280
   1 of 1
   +0
   
7. https://github.com/NVIDIA/cccl_private/issues/209
8. [FEA]: Refactor C++/C dispatch layers to avoid code duplication #2448
   2.8.0 feature request +2
   

[gevtushenko]

cuda.parallel module was introduced to CCCL repo and CI. Further work on cuda.parallel will be tracked by a different issue.