Ensure CMake finds CUDA libraries from host packages

[robertmaynard]

Checklist

• [x] Used a personal fork of the feedstock to propose changes
• [x] Bumped the build number (if the version is unchanged)
• [ ] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• [x] Ensured the license file is being packaged.

Consider a conda recipe that looks like:

requirements:
  build:
    - {{ compiler('c') }}
    - {{ compiler('cxx') }}
    - {{ compiler('cuda') }}
    - cmake >=3.26.4
    - sysroot_{{ target_platform }} 2.17
    - cuda-version =12.0
  host:
    - libcublas-dev
    - libcublas-static
    - cuda-version =12.0

We have placed the CUDA compiler in build as the compiler supports cross-compiling. We place the cublas libraries in host since it is only for the target platform we are building. Now since the compiler is in build it isn't aware of the host libraries and doesn't report those locations as implicit search dirs.

So for CMake to find the cublas libraries we need to extend CMAKE_FIND_ROOT_PATH to search $PREFIX/${targetsDir}.

[conda-forge-webservices[bot]]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[jakirkham]

cc @isuruf

[jakirkham]

Anything else needed here?