Estimating Moderated Causal Effects with Time-Varying Treatments and Time-Varying Moderators: Structural Nested Mean Models and Regression with Residuals