Moved from heat capacity to specific heat
to take into account individual mass of a particle.

The old name (REAL) caused problems when coupling the Trilinos library
since some component has a global define with the same name.

use cuda as a first level language in CMake with `enable_language(CUDA)`
enable tests where cuda and codegen are both activated
various minor fixes

Please use mocks for these kind of tests.
a) deterministic
b) faster

If you are familiar with the coupling feel free to remove the NOLINTs!

I am unsure about const correctness in the python coupling.
If you are familiar with the coupling feel free to remove the NOLINTs!

To remove code duplication the distinction is now moved
into the generator. Default kernels are "with shape".

When processes contain no blocks after refinement the neighbor list
is not correctly reset.