Amazingly, this really does have a measurable effect on the overall computation
speed.  Assembly times for the global Hessian and gradient for the Cosserat shell
energy problem drop by about 5% (!)

[[Imported from SVN: r9837]]

That's obviously more efficient when solving one linear equation
with multiple right hand sides.

[[Imported from SVN: r9836]]

Previously the energy norm was used, but as we are solving non-convex
problems the Hessians may be indefinite, and then the EnergyNorm is not
a norm.

[[Imported from SVN: r9835]]

[[Imported from SVN: r9834]]

[[Imported from SVN: r9833]]

[[Imported from SVN: r9832]]

[[Imported from SVN: r9831]]

[[Imported from SVN: r9830]]

[[Imported from SVN: r9829]]

That makes it easier to keep them consistent.  Also, the code is shorter.

[[Imported from SVN: r9822]]

[[Imported from SVN: r9821]]

[[Imported from SVN: r9820]]

This used to be 'int', because dunepython didn't support the python-to-C
conversion for bool.

[[Imported from SVN: r9819]]

Previously we called hess_mat (vector-mode second derivatives) with the
canonical basis vectors, and then rotated the resulting matrix.
This is wasteful, as more directions are used than necessary.
This patch uses the minimal number of directions.

[[Imported from SVN: r9818]]

[[Imported from SVN: r9817]]

The hasObstacle_ and obstacles_ fields are now single arrays.
The corresponding fields for the entire grid hierarchy is assembled
within dune-solvers now.

[[Imported from SVN: r9816]]

This almost halves the assembly time for the Hesse matrix.  Nice!
Nevertheless I still keep the scalar-mode code around, hidden
behind and #ifdef.  Higher-order vector-mode in ADOL-C appears
to be not quite as reliable as the rest, so I want to be able to
easily compare results with the scalar code.

Also, unlike the scalar code, the vector-mode one again computes
the full Hessian in the surrounding Euclidean space.  Only afterwards,
that full Hessian is multiplied by the tangent space basis vectors
to obtain a Hesse matrix on the tangent space.  To obtain better
efficiency one could have ADOL-C compute the Hessian in the tangent
directions directly.  However, ADOL-C sometimes segfaults... when
doing that.

[[Imported from SVN: r9815]]

[[Imported from SVN: r9814]]

[[Imported from SVN: r9813]]

In particular, I can use math.cos in the initial values.

[[Imported from SVN: r9808]]

And add a little quirk: I don't know how to do 'import math' in dunepython yet.
Therefore I have to replace the cosine in the initial configuration by a
series expansion...

[[Imported from SVN: r9804]]

[[Imported from SVN: r9803]]

[[Imported from SVN: r9802]]

[[Imported from SVN: r9801]]

[[Imported from SVN: r9800]]

[[Imported from SVN: r9799]]

Newer versions of ADOL-C do not define 'overwrite' anymore, hence there
is no risk of macro clashing.

[[Imported from SVN: r9798]]

[[Imported from SVN: r9797]]

[[Imported from SVN: r9796]]

[[Imported from SVN: r9795]]

That way, we can now control what vertices are Dirichlet and Neumann without having
to recompile the code.

[[Imported from SVN: r9794]]

This will eventually allow to not hard-wired the Dirichlet vertices
into the C++ code, and change them without recompiling.  This also
means we can remove some preprocessor switches.

Currently, though, the Python code is still hard-wired, so we are
only half-way there.

Note that this patch may break several test problems.  I did not test
all of them.

[[Imported from SVN: r9793]]

[[Imported from SVN: r9792]]

[[Imported from SVN: r9791]]

[[Imported from SVN: r9790]]

[[Imported from SVN: r9789]]

[[Imported from SVN: r9788]]

[[Imported from SVN: r9787]]

[[Imported from SVN: r9786]]

[[Imported from SVN: r9785]]