Large example data from the medical domain; including more complex dataset transformations

[constantinpape]

In https://github.com/ome/ngff/pull/57 @thewtex remarked that some images may need a rotation transformation to map from data to physical space directly. We decided to not include this for now, but he provided some (large-scale!) example data:

I know that some data acquisitions or processing software may not account for rotation's in the data, but I hope that we can add optional support for rotations in a future update as this metadata is essential in medical imaging or emerging large-scale whole-organism microscopy. Support for rotation's will make a much better case for NGFF in these applications. Example datasets that we can use for testing are:

1. The [Allen CCF Mouse Brain Atlas](https://help.brain-map.org/display/mousebrain/API)

2. The [MNI Human Brain Atlas](http://nist.mni.mcgill.ca/mni-ftd-templates/)

3. A brain image downloaded from [OpenNeuro](https://openneuro.org/).

These images are in formats like the BIDS NIFTI, and contain metadata on how the whole brain is oriented relative to the imaging system. This metadata is critical for researchers that want to work with these datasets. We should be able to, for example, load these datasets in napari and observe that they are rendered with the same anatomical orientation.

This is something we def. want to support in the future (though I think it's still up for discussion how exactly). A good next step would be to provide a concrete (and not too large) example image based on one of the examples above (3 looks promising). Also relevant to this discussion: https://github.com/ome/ngff/issues/74 https://github.com/ome/ngff/pull/63