Move intelligence to Impl::create_mirror*

[pzehner]

This PR aims to simplify the code by moving the duplicated intelligence of create_mirror and create_mirror_view to Impl::create_mirror and Impl::create_mirror_view. The former functions now behave as a pass-trough that only call their Impl:: counterpart.

This PR is part of a of a larger effort which aims to fix kokkos#6842.

A functor CopyUnInit is implemented in the test file core/unit_test/TestViewAPI.hpp which contains a View as an attribute whose type is hard-coded using a MirrorViewType implementation object here (it is the only test in the file containing an implementation object). The type expected by the functor is a view whose Space parameter corresponds to the template Space indicated in the same functor here.

In the develop branch, the overload of the create_mirror function called in the test imposes the following return type: Impl::MirrorViewType<Space, T, P...>::view_type which matches the type expected by the constructor of the functor in the test.

In this branch, the return types of the functions are auto. Therefore, the return value is decided at a single place. The overload of create_mirror_view called calls the function Impl::create_mirror_view, which in turn calls the function create_mirror, which returns the following type: Impl::MirrorType<memory_space, T, P...>::view_type that does not match the type expected by the constructor of the test because of the Space::memory_space passed as a Space parameter to the output View.
It should be noted that in the develop branch, several return values are set, but they are different.

To fix the mismatch of types encountered with this PR, we decided to simplify the test code by removing the hard-coded type and simply deducing the object's type by passing it to the constructor. It should be noted that this functor is not used anywhere else in the code.

Tests:

• Clang 10.0.0 with CUDA 11.0;
• Clang 8.0.0 with OpenMP;
• GCC 8.2.0 8.4.0 with OpenMP;
• GCC 8.2.0 8.4.0 with CUDA 11.0 11.4;
• Intel Classic 19.0.5 2023.2.1 with OpenMP;
• Intel LLVM 2022.0.0 2024.0.2 with SYCL;
• Intel LLVM 2021.1.1 2023.2.1 with OpenMP;
• MSVC 19.29;
• NVHPC/PGI 22.3; and
• ROCM 5.2.0 with HIP.

[cedricchevalier19]

Be sure to run clang-format8.

[cedricchevalier19]

Be sure that auto works properly on all platforms.

[pzehner]

We should run all the different backends with the minimum supported compilers. So far, we could build for Serial, CUDA and SYCL (latest versions of compilers, mostly).