Improve code portability

[jngrad]

Description of changes:

• address compiler diagnostics for GCC 13 to 15, Clang 19 to 21, nvcc 12.0 to 12.8
• skip compilation of deprecated Kokkos and waLBerla features
• bump all dependencies to get full C++20 support
• modernize user guide with Sphinx tabs to wrap build instructions for different toolchains
• redesign runtime error system to use Python's stderr instead of the lower-level std::cerr (which cannot be redirected to a buffer/filehandle/stream at the Python level)