Document that published `cuda-bench` wheels are multi-CUDA
wheels containing both CUDA 12.x and CUDA 13.x native extensions,
with runtime selection driven by the installed cuda.bindings major
version.
Clarify that the cu12/cu13 extras select compatible cuda-bindings
dependency families rather than different cuda-bench wheels.
Update source-build instructions to use CUDAARCHS=all-major, avoid
editable installs while the versioned extension layout is unsupported
there, and note that local builds only produce the extension for the
Toolkit found by CMake.
closes#391
1. Download and include CPM.cmake, version 0.42.0
2. Use CPM.make to get Pybind11
3. Update to use pybind11=3.0.0
4. Also use CPM to configure/build nvbench
Make it explicit in README that we build and locally install NVBench first,
and then build Python package use the library as a dependency.
The nvbench library is installed into Python layout alongside the native
extension.
Edit wheel.packages metadata to include namespace package "cuda".
Updated README to remove the work-around of setting PYTHONPATH,
as it is no longer necessary.
