Migrate to FindCUDAToolkit.cmake

[SomeoneSerge]

Describe the feature and motivation

NVIDIA considers FindCUDA.cmake (find_package(CUDA)) deprecated: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/#importing-tarballs-into-cmake. The common approach to consume cudatoolkit today is to use FindCUDAToolkit.cmake (find_package(CUDAToolkit COMPONENTS cudart cublas ...)).

Aside from FindCUDA being obsolete, there are notable differences between the modules: e.g. FindCUDAToolkit.cmake allows its CUDAToolkit_ROOT variable to take the form of a list of paths, and this way it supports splayed installations of cudatoolkit, where different components (cudart, cublas, ...) may reside in different prefixes (as is the case e.g. in nixpkgs)

Other big sci-comp projects are also moving away from FindCUDA, e.g. pytorch/pytorch#76082

Thanks!

Additional context

No response

[cudawarped]

I think because OpenCV used FindCUDA for both genrating the build files for CUDA code and accessing the CUDA Toolkit it would be difficult to just use FindCUDAToolkit without enabling CUDA as a first class language.

In #23021 and opencv/opencv_contrib#3400 I started the process of moving to CUDA as a first class language which includes using FindCUDAToolkit.

It would be great if you could check if this works for you and if CUDAToolkit_ROOT can except a list of paths when passed to OpenCV.