Scaling issues

[will-moore]

Various scaling issues have been reported. These are due to the initial Scaler designed to scale planes (or a series of planes) in 2D and to work with numpy arrays rather than dask arrays.

• Scaler not scaling all dimensions #262
• Two questions about converting larger than memory ND data into ome-zarr #255 "isn't clear to us if the Scaler() function in ome-zarr-py performs lazy down-sampling."
• Multiscale method in write_images ignored (e.g. nearest vs gaussian) #254

Currently, there are a bunch of different algorithms available for scaling:
All of these generate a full pyramid (List[ndarray]) scaled in X and Y only. Only nearest supports dask arrays.

• nearest
• gaussian
• laplacian
• local_mean
• zoom

Question: Do we want to support n-Dimensional scaling with Dask support for all of the above methods? See example of this at https://github.com/spatial-image/multiscale-spatial-image/blob/efc0ddc167cd2a14ceac4898e4edf08e7789c368/multiscale_spatial_image/to_multiscale/to_multiscale.py#L26

[will-moore]

Also see https://github.com/thewtex/ngff-zarr which does scaling. E.g. https://ngff-zarr.readthedocs.io/en/latest/apidocs/ngff_zarr/ngff_zarr.to_multiscales.html