[C++][Python] Removed references to kinematic model/data in geometry model/data. + fixed unallocated memory map ( due to using reserve keyword instead of resize)

[C++] Removed type from GeometryObject. Still using the enum GeometryType to specify if visuals or collisions must be parsed from urdf file

Conflicts:

	CMakeLists.txt

[C++][unittest] Added neutralConfiguration vector in Model, and fill it when adding joints + test on neutralConfiguration

Renamed JointXXAccessor to JointXX + renamed traits<D>::JointData to traits<D>::JointDataDerived, same for Model.

Conflicts:

	CMakeLists.txt
	src/algorithm/center-of-mass.hxx
	src/multibody/model.hpp
	unittest/CMakeLists.txt

[Minor] Update name in Data of oMof to oMf + in frames algo get directly the value of the parent joint when the relative placement of the frame wrt parent joint is Identity

As it is done in src/multibody/parser/python.cpp, one can create a shared_ptr to a new Model in c++,
send this into a python namespace, update it in python, and continue to work on the Model in c++.

An example to use this as a simple parser is given in unittest/python.cpp and models/simple_model.py

(squashed version)

With next version of Eigen3, it seems that there exists a missing
initialiser when calling the cast operator.

[C++] bodies are now a subpart of the Frames : some Frames are of type BODY and when referring to getBodyId/Name in Model it will look in its frames

[C++] Make distinction between joint and body in Model API ( addJoint, appendBodyToJoint, addJointAndBody instead of addBody)

 [C++][Unittest] Reworked JointAccessor Unittest. tests are now done on every type of joint of the boost variant Model

Completed the visitors on JointVariants for each method of the JointBase interface  and moved it to a different file.

[C++][Python][Major] Replaced the way collisions and visuals are stored in GeometryModel. Now it is two lists of GeometryObject (one for collisions, one for visuals) aggregating information instead of multiple vectors. Now handle multiple collisions/visuals for one link when parsing urdf file and keep track of the mesh absolute path. Changed binding accordingly.

[C++][FixBug][Unittest] Fix bug in calc_aba (bug with boost:fusion preventing a proper acces to jmodel.axis) + reworked unittest for joint prismatic unaligned

[C++][FixBug][Unittest] Reworked unittests for joint revolute unaligned + Fix bug in calc_aba (bug with boost:fusion preventing a proper acces to jmodel.axis)