xenium dataset #2

Todo and caveats for the support of xenium data.

Todo:

• most important point: replace bioformats2raw with dask-image (as it is done in other datasets). In this way the ugly function _ome_ngff_dims_workaround() can be removed.
• the points (transcripts) should include all the columns, not just the transcript type; waiting for the io to work with geopandas (replacing the current anndata implementation)
• test visualization with napari
• suggested by @ivirshup: better approach to the DEBUG variable
• indexes for the annotating table don't match the cells (one start from 0, the other from 1) #11

Old todos (removed):

• the image morphology.ome.tif seem to contain time information (it is converted to tczyx and has multiple t channels). Check with Fiji if this is the case or if it is the result of an error in conversion with bioformats2raw. EDIT: we are not parsing this until we have a use case.

Caveats:

• The table can annotate only one Regions element, while here there are multiple regions available (and also labels could be created to rasterize the polygon masks). We have to decide how to proceed, but this is for later.