Cleanups and simplication of tetrahedron hot path
These are some patches for the simplification of the code and to increase the performance. I reduced a level of indirection by swapping out std::list for std::vector in a small part of the code. Run time for meshing decreased (1min 20 to 1min 12) on my machine for an engineering style model with around 2 million vertices (second order tets).
Would you be able to please confirm these findings?
If confirmed, there are a few other places where std::list can be removed for other easy performance gains.