allows to export BOUT_TOP for all projects

In case of fci, we need to apply parallel boundaries as well.

The for loop took a copy, rather then the field itself.

allows to  compile with cmake

This reverts commit fc3fe5d.

DC is inherently unsuitable, as it depends on the free parameter of the
grid generation, namely all parallel slices can be abitrarily rotated in
poloidal direction, thus at most 1D averaging could be used.

Works, no reason to average

* Always use floor'ed fields
* some additional calls to communicate + parallel BCs

Can be done with boutcore in post processing

This prevents overwriting the boundary conditions of Te etc for the
kappa_{i,e}par, thus probably more correct then the previous version.

Unify forward + backward, only difference is the sign in a few cases.

Add basic CI workflow

* Fix typo (parallel_sheath**s**)
* Suggestions from v5-auto-updater

* Use correct terms for laplacian newsolver
* Cleanup input files

simple low order

Probably not correct. We can implement something correct later - it
should not be of major important.

If we apply standard boundaries, we probably also want FCI ones

Cleanup + boundary conditions