Support adding multiple layers with different voxel sizes separately

[frcroth]

When exploring storage.googleapis.com/neuroglancer-fafb-data/fafb_v14/fafb_v14_orig with voxel size 4,4,40 and then adding storage.googleapis.com/neuroglancer-fafb-data/fafb_v14/fafb_v14_clahe with voxel size 8,8,40 a warning occurs and the resulting data is messed up. What should happen is that on the second dataset, the mags are adjusted so that it starts at mag 2,2,1.

#6670 already implemented this when exploring a data source with multiple layers of different voxel sizes (e.g. segmentation and image), however this does not work when adding one layer and then adding and exploring another layer.

[philippotto]

With #6748 the restriction of uniform voxel sizes might be lifted (might take a few iterations). Therefore, I would advise to not tackle this issue here before it's clear whether the other issue will automatically solve it.

[normanrz]

With #6748 and #6943 merged, we should be able to tackle this now?

[philippotto]

Yes, this should be possible now. Instead of the originally suggested "renaming" of the mag layers, the mag linking could be 1:1. Then, the layers with differing voxel sizes would need an affine coordinateTransformations property which does the scaling. The scaling itself can generalize over arbitrary voxel sizes (doesn't depend on power-of-two-factors).

We still have the concept of a singular "voxel size" property for the entire dataset. I think, the first layer that is added could dictate that voxel size and all other layers get a transformation if their voxel sizes differ.

[frcroth]

@philippotto I think the problem is that the frontend calls exploreRemoteDatasource, but the backend has no knowledge if this is for a new dataset or if it is a new layer for an explored dataset. Thus the backend does not know the voxel size (for which coordinate transformations could then be applied). I think the frontend needs to pass a preferred voxel size on the exploreRemoteDataset API call.