Revisiting Weighted Strategy for Non-stationary Parametric Bandits and MDPs