[FEATURE] Larger than memory images: performant and scalable distributed implementation for workstations and clusters

[GFleishman]

Is your feature request related to a problem? Please describe.
Native cellpose cannot handle images larger than gpu memory or system memory.

Describe the solution you'd like
A performant and scalable distributed implementation that can run on workstations or clusters, processes the data in blocks in parallel or in series, and stitches the results back together.

Describe alternatives you've considered
A few attempts have been made at this in the past. This is the most fully featured and thoroughly tested.

Additional context