There is way more reasons to interpolate a function into the
solution vector and we cover them with one code path:
* Dirichlet boundary conditions (strong ones)
* Initial conditions for instationary expressions
* Initial guesses for Newton solvers

The visitor state regarding indices needs to be correctly reset
for child iteration in these cases.

These might sound completely silly, but they are useful when writing
code generic w.r.t. the number of components

It has single precision support...

will be superseded by our actual model anyway.

We run ut of PIDs during a typical run on KNL.

Additionally, uses the right number of subrefined intervals in the blockstructured case

This reverts commit c33945c3.

Still not beautiful...