Causal Imitation Learning Under Measurement Error and Distribution Shift