reader chunk_dims are too coarse-grained for dask

[toloudis]

Feature Description

Allow more specificity in requested chunk_dims for get_image_dask_data (_read_delayed).

Use Case

Dask best practices for chunk sizing: https://docs.dask.org/en/latest/delayed-best-practices.html
Currently bioio only allows you to choose a dimension AXIS for chunking and always chunks with the whole range of that dimension.
When reading a large file with something like t=500, c=2, z=150, y=1000, x=2000, (each xy slice is about 4MB) we don't really have the option to chunk by just a few, or even half, of the z slices.

Solution

Not sure about the api to use here, but possibly pass in a chunk_size with actual numeric values.
So the user would probably do this, as an example:

im = BioImage(path)
dims = im.dims
im.get_image_dask_data(chunk_size=[1,1,1,dims.Y*0.5, dims.X*0.5]). # (Defaults to something sensible?)