Change control architecture to work in horizontal frame with reference quantities instead of relying more on estimated quantities

Add possibility to use estimator with perfect height and linear velocity estimation (when in simulation)

Remove first column of gait matrix that contained the duration of each gait phase. Now gait matrix has one line for each timestep.

New organization and bidings to call Gait, FootstepPlanner and FootTrajectoryGenerator directly from Python

* Code cleaning and refactor
* Add Types, FootTrajectoryGenerator, Gait, FootstepPlanner classes
* Rename quadruped-reactive-walking qrw

Authored-by: Fanny Risbourg <frisbourg@laas.fr>

Reswitch to old estimator values + Put back things that should not have been cleaned + Try fixing gepetto viewer

Improved initialisation that takes into account the initial position of feet and the initial height of the base based on q_init

Add logging + Rework control architecture (nothing feedbacks from the WBC, everything is taken from the estimated state)

Changing data going from MPC to WBC + Integration of reference velocity to get reference position (starting at 0 0) + Testing different kinds of feedback for the control loop