The Production of Co-Speech Iconic Gestures: Empirical Study and Computational Simulation with Virtual Agents