Support for label images / bitmask segmentations

[keller-mark]

User story

For supporting label images, in Vitessce we use the standard viv.MultiscaleImageLayer (code) combined with a custom renderSubLayers (code, more code, utils) and custom shaders

This code can be upstreamed into Viv but note that currently things like expression-based and obsSet-based coloring are currently fairly tied to the Vitessce data model. The utilities+shader in particular compress this data a lot and support things like out-of-order and non-consecutive observation IDs.

Preferred solution

Possible alternatives