Change default computation of mean plane (and thus of the parametrization) of plane surfaces, to make
it more robust with respect to perturbations of the boundary.

This is crucial for sensitivity analyses and GFace::relocateMeshVertices(): after perturbation of the 
boundary, we want a parametrization of the surface that is "close" to the original one. This cannot be
guaranteed by the automatic SVD-based algorithm.

Unfortunately the new algorithm is also not guaranteed to work in all cases -- but unless we explicitely
prescribe the parametrization of plane surfaces (at least the normal) this is better.

I will revert this as soon as we have something better.

prudently put between #if 0 #endif Those should allow
to produce better quality quad meshes and more robust
triangular meshes (random factor has disappeared) : 
Technically, the divide and conquer algo has been 
replaced by a bowyer-watson with robust predicates.

spallart almaras is kinda working

--no commit message

live with rotations. Two examples were added
in the benchmarks/3d repository

gmsh: Added functions to nodalBasis for dg to compile. (I Think) those functions should be there and not in polynomialBasis, as they are general to any base.
dg: switch from polynomialBasis to nodalBasis

not yet finished ... mesh smoothing has to be adapted 

cross field computataion ok in 2D
high order tools bugs fixed
wing-spline is now a correct geometry
meshing in two passes is now permitted (like in bamg)

- fixed MED node ordering
- polished 64 bit Cocoa Mac OS X FLTK version
- updated copyright

FIXME: please remove dependencies to Geo/ code in Numeric/ !

--no commit message