[[Imported from SVN: r4067]]

bugfix: we need the derivative of the geodesic interpolation in the gradient computation (chain rule), not the derivative of the velocity interpolation

[[Imported from SVN: r4066]]

for debugging: add a method to compute gradients of local geodesic fe functions by finite-difference approximation (only 2d)

[[Imported from SVN: r4065]]

[[Imported from SVN: r4062]]

implement interpolation of the gradient for members of SO(3), some of this is still educated guesswork, though

[[Imported from SVN: r4060]]

[[Imported from SVN: r4057]]