Hybrid-Diffusion Models: Combining Open-loop Routines with Visuomotor Diffusion Policies