ViewFusion: Learning Composable Diffusion Models for Novel View Synthesis