[C++][Frame] Moved the generation of random string in a new file. Modified the default constructor of frame to produce unique frame (as long as the random strings generated are unique)

Conflicts:

	CMakeLists.txt
	src/python/algorithms.hpp

[C++][Python][parser] Added operational_frames in Model as a vector of Frame {name, parent id, placementWrtParentJoint} and exposed it in python. When parsing a fixed joint, a frame is added to the model with the name of the fixed joint

By default, one can now build the benchmarks from make timings. They
are build along with the lib if BUILD_BENCHMARK is set to ON.

Conflicts:

	CMakeLists.txt
	src/python/algorithms.hpp

[C++][Python][parser] Added extra_frames in Model as a vector of Frame {name, parent id, placementWrtParentJoint} and exposed it in python. When parsing a fixed joint, a frame is added to the model with the name of the fixed joint

[bench][unittest] Reworked the comparison test between pinocchio and hpp : more tests for benchmarks along with functional tests ; removing magic paths to urdf models + modification of romeo urdf model. Also added a specific BUILD_TESTS_WITH_HPP cmake variable. If you want to run tests that execute hpp robot loading, be sure to have the package romeo_pinocchio installed (it is just a package done done by hand for the moment)

[C++] computeDistances and computeCollisions moved out of GeometryData. Added a vector of boolean handling the state of whether or not the pairs are in collision in GeometryData

[C++][CMake][Doc] Split geometry parser into two parts.Added guard to file to avoid multiple redefinition

[CMake][Python] Exposed GeometryModel and GeometryData in python. Little reworked of the headers_variables in the cmakelists

[benchmark] Added benchmark of atomic computations with geometry objects such as updating global positions, determine if two objects are in collision, or compute the distance between two objects

May need some change according to the current version of the
architecture

[Python] Binding the different joint models. The exposure is done through the variant. In joints-models, you can specialise the binding for any jointModel you want.

  * Implement exp3, exp6, log3, and log6 functions by copying
    implementation from explog.py.
  * Add same documentation as in explog.py.

[Major][C++] Added Joint limits to joint model and to Pinocchio Data. Also added an algorithm to compute joint limits. Based on the model of the limits, it computes and store the actual limits in Data. We then have access to the values in Data in python binding. The algorithm is here in future cases where the limits will not only be static but dependent of the configuration