[BUG] Document or expose the 'remove big masks' filter

[ashishUthama]

Describe the bug A user (the40x) struggled with segmenting their large mask. The cell probability output looked clean and they were puzzled why despite tuning flow error/prob threshold etc, they were unable to get the final mask out Link: https://forum.image.sc/t/cellpose-failing-to-produce-masks-despite-obvious-flows-and-cellprobs/86849/4

Screenshots Above post has detailed images.

They eventually traced this down to: https://github.com/MouseLand/cellpose/blob/5fb74fde155c35dafcd096fd66118b3cfdd174a0/cellpose/dynamics.py#L713C3-L713C3

[ashishUthama]

I could not find anything relevant with a quick search for 'max' in the online doc, I might be mistaken and this may be documented somewhere.

There likely was a good reason to add this. So maybe we could expose this as max_size and default to the same 0.4 value?

[mrariden]

I think exposing this via the CLI/API makes the most sense. @carsen-stringer