Differential Privacy for Clustering Under Continual Observation