Add method to create masks grouped by label id

[dominikl]

Add a method to create masks from labelled z-stack image, grouped by label ids. These can used to create "3D'ish" ROIs. Input is a list of planes so that the output from the getPlanes() method can be used directly.

[sbesson]

The ability to transform multidimensional labelled image into a series of ROIs with all masks for a given label alongside a dimension is a use case we keep encountering. One small downside of the approach proposed in this PR is that it makes strong assumptions about the planes and a new method would be required e.g. if one wanted to group masks by label across the time dimension.

If the main objective is to group all masks by label for a listof planes, another option might be to use the return value of mask_from_label_image. As this API returns an ordered list of masks, where the list index is effectively the label value, it might be possible to combine it with Python's built-in zip/itertools.zip_longest functions e.g.

from itertools import ziplongest

masks_per_plane = []
for z, plane in enumerate(planes):
    masks = masks_from_label_image(plane, rgba, z, c, t, text, raise_on_no_mask=False)
    masks_per_plane.append(masks)

masks_per_label = list(zip_longest(*masks_per_plane))

Note the example above will fill missing values with None by default. For consistency, it might be preferable to fill it with empty masks.

[dominikl]

Indeed, the index of the masks coming from masks_from_label_image is basically the label. I rewrote the method accordingly, which makes it easier. I'm not sure about ziplongest... I rather write a few lines more code, but therefore make it a bit easier to understand what's actually going on.

[sbesson]

As discussed earlier today, let us drive this functionality with the IDR submission as a concrete use case while we make progress on the API implementation and tests.