DataArray encoding needs to be deleted in order to save a dataset

[eric-czech]

We need some way for users to handle this situation (i.e. pydata/xarray#4380):

ds = sg.simulate_genotype_call_dataset(10000, 500).chunk(chunks=100)
del ds.attrs['contigs']
ds.to_zarr('/tmp/ds.zarr', mode='w')
xr.open_zarr('/tmp/ds.zarr').chunk(chunks=150).to_zarr('/tmp/ds2.zarr', mode='w')

NotImplementedError: Specified zarr chunks encoding['chunks']=(100,) for variable named 'variant_contig' 
would overlap multiple dask chunks ((150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 
150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 
150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 150, 
150, 150, 150, 150, 150, 100),). This is not implemented in xarray yet. Consider either rechunking using 
`chunk()` or instead deleting or modifying `encoding['chunks']`.

The .encoding['chunks'] value is stale in this situation and can be ignored.

[tomwhite]

Fixed in #392