3d -> 9 fma + 18 mul (without -1 *)
2 fma more than C++, 4 mul/add more than C++

3d -> 5 fma + 4 mul (without -1 *)

Matrix inversion at code generation time does only work to
a very limited extent (up to n=4). We can instead assemble
the tensor in C++ and invert it there (e.g. using Dune::FieldMatrix)

Lets see in the future what is necessary and whats not