Iterative Residual Policy: for Goal-Conditioned Dynamic Manipulation of Deformable Objects