Spawn one instance of robot_state_publisher by human, to publish the TF frames based on the joint state

For 3D skeletons, default URDF is published + default joint state.
Need to wire up URDF adjustment + IK to compute joint state

This follows the ROS4HRI convention

Will be ultimately used to generate on the fly URDF models
of humans, with dimensions adjusted to each human

Currently, can only subscribe to image topic.

TODO:
- use camera calibration data if available
- publish the TF transforms of each joint
- generate and publish the URDF model of the human(s)