DIRIGENt: End-To-End Robotic Imitation of Human Demonstrations Based on a Diffusion Model