The implemented formula was actually for the projection of Stiefel manifolds, of which
the orthogonal matrices are only a special case.  For this special case, one term
drops out and doesn't have to be implemented.

... with values expected to be given in the parameter file.

Scaling the trust region is the key to success, because in general the midsurface
deformation and the rotations will have values in different ranges.

This marks the last step of porting the 'cosserat-continuum' application away
from the dune-fufem bases.  One must not(!!!) mix dune-fufem bases with
dune-functions bases, as they use different dof orderings at least in the P2 case.

Because this object doesn't exist in dune-functions bases anymore.

[cleanup] Move the FD approximation of the GFE function gradient from the GFE function into the test

It is a part of the test harness, and should not clutter the implementation of the
GFE functions themselves.