Consider using nuclei locations for Xenium data when cells_as_circles is True

[LucaMarconato]

For Xenium data we have a parameter cells_as_circles to improve the speed of visualization by representing cells as circles instead of polygons.

Since the Xenium boundary segmentation is computed as the maximum expansion of the nuclei segmentation, using the cells radii would lead to a cluttered view where cells overlap on each other. So currently we are using the nuclei radii. This can be a bit confusing. Also, the nuclei centroids do not correspond to the cells centroids.

We should therefore consider one of the following.

• Keep using the cells centroids and the nuclei radii
• Use the nuclei centroids and nuclei radii
• Use the nuclei centroids and nuclei radii with a new parameter nuclei_as_circles and then use the cells centroids with cells radii with cells_as_circles

I think the last approach is the best.

[LucaMarconato]

See this: #173

Proposed fix:

• use the cell centroids and cell radii (stop using the nuclei radii as they can be missing).
• also, by default assign the table to the cell labels (better visualization performance, even better than circles, but not than points)
• add in the docstring a note saying that the table can be assigned to circle using set_table_annotates_spatialelement(); and put a code example, either in the docstring of xenium() or of set_table_annotates_spatialelement().