ENH: 2D chunks for all

[jklymak]

This allows 2D chunking for all mds files, not just the LLC cases.

This is a total hack at this point/proof of concept. However, it works for me, and allows me to get 2-D chunks and parallel work on a machine that doesn't have enough memory for my model run (for some reason I can't discern - each variable is only 5 Gb).

• needs docs
• needs tests
• probably needs to be properly interested with the other chunking options.

with xm.open_mdsdataset(data_dir, prefix=['means'], 
                                endian="<", geometry='cartesian',
                                chunks="2D") as ds:
   print(ds)

Here Nz=400:

...
Data variables:
    VVEL     (time, Z, YG, XC) float64 dask.array<chunksize=(1, 1, 1216, 1440), meta=np.ndarray>
    UVEL     (time, Z, YC, XG) float64 dask.array<chunksize=(1, 1, 1216, 1440), meta=np.ndarray>
    THETA    (time, Z, YC, XC) float64 dask.array<chunksize=(1, 1, 1216, 1440), meta=np.ndarray>
...