This actually enforced writing of three tape files of ADOL-C (see taping.h)!

Catch an exception that might be thrown by the localEnergy_ and solve the problem again using a smaller trust region.
Also make sure the ADOLC-Tape gets closed properly

Currently, all implementations of variational problems inherit
from LocalGeodesicFEStiffness, which requires gradient and Hessian
implementations.  However, most of these implementations really
only implement the energy, and leave computations of first and
second derivatives to algorithmic differentiation.  To make
this clearer, this patch introduces an abstract base class
LocalEnergy for all implementations that really only have
an energy.

When using ADOL-C together with MPI we need to use different tags in the
ADOL-C function calls for each process. If we use the same tag, then the
processes will interfere with each other causing undefined behaviour.

This is how the new dune-functions bases should be used.  Also, it is yet
another step towards merging CosseratEnergyLocalStiffness and MixedCosseratEnergy.

[[Imported from SVN: r10103]]

Conceptually cleaner, and makes for simpler code.  Most importantly, we will want
to pass the Basis::LocalView to the local assemblers.  For that we need the actual
basis, not just the grid view and local finite element.

[[Imported from SVN: r10082]]

[[Imported from SVN: r10068]]

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]]

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]]

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

[[Imported from SVN: r9798]]

Instead, only let ADOL-C evaluate the Hessian in the directions
of the vectors of the orthonormal frames.  The result is the same,
but we get away with fewer calls to ADOL-C, which, in my Cosserat
shell example, shaves off about 10% of the assembly time.

[[Imported from SVN: r9779]]

[[Imported from SVN: r9572]]

[[Imported from SVN: r9571]]

Because the gradient is needed anyway when computing the Hesse matrix.
Not computing it separately a second time shaves off about 8% wall time
of each iteration.

[[Imported from SVN: r9555]]

The Weingarten map is block-diagonal for a product space.

[[Imported from SVN: r9517]]

[[Imported from SVN: r9422]]

[[Imported from SVN: r9415]]

Because it may contain a projection onto the manifold, which
is part of the energy evaluation if we view the energy as
being defined on all of R^n (as ADOL-C does).

[[Imported from SVN: r9414]]

Not actually working yet, though.

[[Imported from SVN: r9412]]

[[Imported from SVN: r9395]]