Add typing to `img_group_model`

[chungmin99]

Currently, it's unclear what the dimensionality + type of the call function should be!

Need to define this before adding support for additional grouping types.

In particular:

• the resolution of the mask should equal resolution of the image
• the mask should be boolean (not 0/1s)
• the masks should be ordered in increasing 2D size