[3D] Rethink 3D problem with tetrahedra

[th-skam]

Changes and reasoning

• For 3D, we now solve using an iterative solver, parallel force and derivative accumulation in the LinearSmallStrainFEMForceField SOFA component
• I switched the scheme to do more iterations on the Newton Raphson, even though it's a linear problem. This allows you to have few iterations of CG on coarse meshes but on fine meshes you let the Newton iterate on the residual.
• I also added a quadrature rule for tetrahedra that does not get affected by the direction of the diagonal.
• I switched from a 3-2-1 Rigid body constraint BC to 1 face carrying Dirichlet and 5 faces having a Neumann (traction) BC. The convergence rates are okay (there will be some non-exactness (e.g. = 1.97 for L2) but remember we are using an iterative solver).