[ISSUE] Remove index_vector from Grid. It is rather ugly... Also value traits
are missing in a lot of the routines. Code coverage is pretty bad. Thus there is
no way of telling which routines are actually working.

[ISSUE]  utility/image_analysis: Should use math_types and value_traits, but it
does not! Besides the library is not used anywhere in the code. Thus we have no
way of knowing if it still works?

[ISSUE] Should we drop core/containers/hash_map?

[ISSUE] core/geometry/t4_gpu_scan: Should we add a namespace?

[ISSUE] core/geometry/t4_cpu_scan: Should we add a namespace?

[ISSUE] Use value traits in FEM library, replace unsigned int with size_t
wherever appropriate.

[ISSUE] core/math/math_polar_decomposition.h introduces its own sub namespace polar_decomposition

[ISSUES] Following drawbacks of the MBD library has been reported:

- State information in rigid bodies and nodes are not cleared when reusing rigid
bodies from a previous configuration. This causes unwanted effects in caching
(and like non-zero velocities eventhough no velocities are set on
initialization).

- Use value traits everywhere

- Maybe move joint stuff from bodies into edges

- Refactor bodygroup to only include bodies and edges. Thus create a two-stage
contact and constraint iterator on a body group, which in the outer-stage
iterates over edges, and in the inner state iterates over contacts or
constraints.

- STL Heaps suck, their performance is really bad, it would be nice to replace
them with something else. See in sequential collision resolvers

- The util folder contains drawing and mel support! This really do not belong in
a general purpose MBD simulator framework, maybe we should kill it?