JK/mesh

Closes #69 and #70

See merge request !96

Despite the ugly diff, the code did not change much - only the two
static class member functions became plain functions, changed order and
indentation level, and the struct DecomposeMesh was removed.

It calls Array::getView which is not __cuda_callable__. It is used only
from the host anyway.

fixes #69

Before this change, the error was:

XMLVTK error: failed to parse the file as an XML document.

fixes #70

Loops

See merge request !94

The kernels are not interesting for users and Doxygen made a mess out of
various CUDA declarations. The 'detail' namespace is considered internal
and excluded from processing by Doxygen.

- passing indices and array/vector values by reference is useless
- declaring arguments passed by value as "const" is useless
- using "IndexType" and "ValueType" is better than "Index" and "Value",
  because Doxygen creates an automatic link to the description of the
  type
- various other documentation and whitespace fixes

It is an implementation detail specific to the CutMeshFunction class,
not something that should be part of the public interface. There are
many problems with this approach: e.g. it does not scale to higher
dimensions and it is not parallel.

- interface renamed from `TemplateStaticFor` to `staticFor`
- also added tests and updated documentation, read it for details

The main purpose of this algorithm is to do loop unrolling, not to
handle static/constant bounds. The true StaticFor will handle the
iteration indices in such a way that they can be used in constant
expressions (e.g. passed as template arguments).