I don't see an elegant way to make this work for a P3 space, even on a single
processor.  Since I need P3 measurements real soon now, I just disable the
entire code.  Not nice.

[[Imported from SVN: r10047]]

The multigrid hierarchy is built in two different ways, depending on whether
we use a P1NodalBasis for the discretization, or anything else.  Previously,
this switch was made using the preprocessor.  The patches changes the code
to use std::is_same instead.

[[Imported from SVN: r10046]]

[[Imported from SVN: r10037]]

This is much cleaner than handing over the Grid only, and then trying
to guess the basis using the preprocessor.  After all, the
RiemannianTrustRegionSolver calls the assembler, so it may know the
function space basis.

[[Imported from SVN: r10035]]

By introducing a factory class that produces GlobalMapper and LocalMapper
types for a given dune-fufem-style function space basis.

[[Imported from SVN: r10034]]

[[Imported from SVN: r10033]]

[[Imported from SVN: r10032]]

[[Imported from SVN: r10031]]

... while keeping compatible with dune 2.3

[[Imported from SVN: r10030]]

This is actually what is claimed in the paper.  Previously, the mirror
axis was y=0.005.  Having the grid symmetric across the x-axis itself
simplifies the expressions for the boundary conditions a little bit.

[[Imported from SVN: r10029]]

Kudos to Carsten Gräser for those new Python functions.

[[Imported from SVN: r10028]]

[[Imported from SVN: r10027]]

[[Imported from SVN: r10026]]

[[Imported from SVN: r10025]]

This implements a global GFE function (inheriting from VirtualGridViewFunction).
However, the return value of this function is _not_ a manifold class.  Rather,
the manifold is embedded into a Euclidean space.  Hence this class is formally
simply vector valued, and can be used in error measurement routines and such.

[[Imported from SVN: r10024]]

This is only a prototype, because the GFE solution function is replaced by
on where the interpolation is done in the embedding space.  That defeats the
purpose of measuring the GFE discretization error :-) but it is easier to
implement.  The real thing will follow shortly.

[[Imported from SVN: r10023]]

[[Imported from SVN: r10022]]

We can do it without the preprocessor.

[[Imported from SVN: r10021]]

I don't think I will need that again in the foreseeable future.
If ever it should be needed again the code can be taken from the
subversion archive.

[[Imported from SVN: r10020]]

The plan is to minimize the harmonic energy starting from the inverse
stereographic projection.  That projection already is the minimizer
of the energy, so it should be possible to measure convergence orders
against a known solution.

[[Imported from SVN: r10019]]

Do not fall back to 'harmonicmaps.parset' by default.

[[Imported from SVN: r10018]]

[[Imported from SVN: r10017]]

[[Imported from SVN: r10016]]

[[Imported from SVN: r10015]]

[[Imported from SVN: r10014]]

[[Imported from SVN: r10013]]

[[Imported from SVN: r10012]]

[[Imported from SVN: r10011]]

... so much difficult code obsolete now... AD rocks...

[[Imported from SVN: r10010]]

[[Imported from SVN: r10009]]

[[Imported from SVN: r10008]]

This is much more flexible and *much* easier to use.

Along the way, this patch scraps several old hard-coded Dirichlet boundary
values.  I don't even remember what they were good for.  If they are needed
again they should also be reimplemented in Python.

[[Imported from SVN: r10007]]

According to Christof Melcher, this produces a field of topological
charge q=1  (or -1, not sure).

[[Imported from SVN: r10006]]

[[Imported from SVN: r10005]]

[[Imported from SVN: r10004]]

[[Imported from SVN: r10003]]

[[Imported from SVN: r10002]]

[[Imported from SVN: r10001]]

[[Imported from SVN: r10000]]

[[Imported from SVN: r9999]]