UMO: Unified In-Context Learning Unlocks Motion Foundation Model Priors | Signal Canvas | ScienceToStartup