[JTC] Fix hold position mode with goal_time>0

[christophfroehlich]

• In hold position mode with goal_time>0: If the JTC can't hold the position (wrong PID tuning, effort limit,..), then it might enter a loop setting a new hold position again and again. Fixes [JTC] Hold position & goal_time>0: Infinite loop #757
  I had to introduce a new class variable to know if we are in holding mode, which breaks ABI. This is also used for [JTC] Add time-out for trajectory interfaces #609 to avoid deadlocks with timeouts.
• I improved the test criterion for the cases with no position command interface: We can check only with active PID mode if the correct position is held.
• set_hold_position generated invalid trajectory messages (use validate_trajectory_msg(current_pose_msg); to test that). I changed that to have the correct fields after sampling (with spline and also none), acc. to the docs of the trajectory representation
• set_hold_position should be now realtime-safe: there is no reason to do memory allocation now where DOF and interface types are fixed

[mergify]

This pull request is in conflict. Could you fix it @christophfroehlich?