Fix iCubGui visualization during YOGA++ demo

[traversaro]

The iCubGui visualization is shitty during the YOGA++ demo, for a number of reason:

• For robot models generated using the v2 generation workflow directly from Creo model, the frames used by the iCubGui for visualizing contact forces are not present (except for l_foot_dh_frame and r_foot_dh_frame) : https://github.com/robotology-playground/icub-model-generator/issues/11 .
• The iCubGui has a long standing bug for which forces under the floor are not correctly visualized : https://github.com/robotology/icub-main/issues/29 . We typically used the workaround of visualizing the robot levitating 1 meter over the ground to overcome this issue.
• The wholeBodyDynamicsDevice is not publishing the floating base position estimate, as the old wholeBodyDynamicsTree did. The rationale is that logic could be moved to another device specifically designed for odometry, that could as well be loaded in the pc104. The class implementing the computational part is already implemented in http://wiki.icub.org/codyco/dox/html/idyntree/html/classiDynTree_1_1SimpleLeggedOdometry.html#a189c13e9ff5d64f3b8b05034f7cedb64 , we just need to write the yarp device (probably recycling some of the code from the wholeBodyDynamicsDevice).

[iron76]

A nice visualisation is fundamental for presenting our work to daily visitors. Do we have an estimate of how much effort is needed for this issue to be addressed?

[traversaro]

The floating base estimator bit should be fixed by https://github.com/robotology/codyco-modules/pull/197 . For the DH frame stuff, it would take some more time, probably two/three days.

[traversaro]

A possible long term solution is just have a tool similar to iCubGui in the sense that it reads robot configuration and robot external forces from existing YARP ports, but that is able to load directly URDF models, reusing the work that has been already done in the iDynTree Visualizer C++ library, see https://github.com/robotology/idyntree/issues/837 .