Materialization at selected MIP levels

[RussTorres]

A common use case (e.g. MICrONS dataset) is to render out an aligned dataset at MIP1 (2,2,1 downsample) for further processing (dense alignment and segmentation) as well as a higher downsampling level for volume previews.
Additionally, writing these to a chunked volumetric dataset like neuroglancer precomputed or zarr is important to most workflows.