Add export targets for FetchContent

[SeverinDiederichs]

This PR proposes to export the libraries core, cpu, and cuda to make them available for FetchContent builds.
This eases the integration into other projects and avoids to use the

# Hack for people using the disgusting mal-practice of pullling in external
# projects via "add_subdirectory"...
if(NOT PROJECT_IS_TOP_LEVEL)
    add_library(covfie::core ALIAS covfie_core)
endif()

With this PR, integration of covfie into AdePT by fetching it from source is as easy as:

    include(FetchContent)
    FetchContent_Declare(
      covfie
      GIT_REPOSITORY ${ADEPT_COVFIE_REPO}
      GIT_TAG ${ADEPT_COVFIE_BRANCH}
    )

    set(COVFIE_PLATFORM_CUDA ON CACHE BOOL "Enable CUDA support in covfie" FORCE)
    FetchContent_MakeAvailable(covfie)

[stephenswat]

Hi Severin, thank you for the PR! Could you please elaborate what problems you face if you don't add this line? I think we are using the library succesfully through FetchContent in detray already, so I'm curious to know what problems you are facing and why.

[SeverinDiederichs]

Just a summary, as we discussed this offline:

Without this PR, it the CMake configuration in AdePT fails with:

CMake Error in CMakeLists.txt:
  export called with target "AdePT_G4_integration" which requires target
  "covfie_core" that is not in any export set.

As far as I can tell, the reason is that covfie currently exports via
install(TARGETS covfie_core EXPORT ${PROJECT_NAME}Targets),
which registers covfie_* into the targets only for installation.

Using export(EXPORT ${PROJECT_NAME}Targets FILE "${CMAKE_BINARY_DIR}/${PROJECT_NAME}Targets.cmake") exports covfie_* such that it is registered already during the CMake configuration step. This way, it is available in FetchContenteven without an installation.
(This is my current understanding of the issue)

[stephenswat]

Thanks a lot for the PR!