Support for runtime Pose parameter

[ZdenekM]

At the moment, all poses that are used as parameters come from Action Points (pose = point's position + one of its orientations). However, one can't create a program that would e.g. call an action of some object detector returning Pose and then use it as a parameter for some consecutive action (e.g. a robot's pick action). We should create a new parameter type, which will, however, be supported only for referencing previous results (otherwise it would duplicate the functionality of Action Points).