Add testing_data with arbitrary chunking

[tcompa]

I run scripts 0 and 1, as a setup, and then rechunk all multiscale levels

$ python rechunk_zarr.py

old_array=dask.array<from-zarr, shape=(3, 1, 2160, 5120), dtype=uint16, chunksize=(1, 1, 2160, 2560), chunktype=numpy.ndarray>
new_array=dask.array<rechunk-merge, shape=(3, 1, 2160, 5120), dtype=uint16, chunksize=(1, 1, 50, 50), chunktype=numpy.ndarray>

old_array=dask.array<from-zarr, shape=(3, 1, 1080, 2560), dtype=uint16, chunksize=(1, 1, 1080, 2560), chunktype=numpy.ndarray>
new_array=dask.array<rechunk-merge, shape=(3, 1, 1080, 2560), dtype=uint16, chunksize=(1, 1, 50, 50), chunktype=numpy.ndarray>

old_array=dask.array<from-zarr, shape=(3, 1, 540, 1280), dtype=uint16, chunksize=(1, 1, 540, 1280), chunktype=numpy.ndarray>
new_array=dask.array<rechunk-merge, shape=(3, 1, 540, 1280), dtype=uint16, chunksize=(1, 1, 50, 50), chunktype=numpy.ndarray>

old_array=dask.array<from-zarr, shape=(3, 1, 270, 640), dtype=uint16, chunksize=(1, 1, 270, 640), chunktype=numpy.ndarray>
new_array=dask.array<rechunk-merge, shape=(3, 1, 270, 640), dtype=uint16, chunksize=(1, 1, 50, 50), chunktype=numpy.ndarray>

old_array=dask.array<from-zarr, shape=(3, 1, 135, 320), dtype=uint16, chunksize=(1, 1, 135, 320), chunktype=numpy.ndarray>
new_array=dask.array<rechunk-merge, shape=(3, 1, 135, 320), dtype=uint16, chunksize=(1, 1, 50, 50), chunktype=numpy.ndarray>

And here is the content of 20200812-CardiomyocyteDifferentiation14-Cycle1_mip.zarr/B/03/0/0/.zarray:

{
    "chunks": [
        1,
        1,
        50,
        50
    ],
    "compressor": {
        "blocksize": 0,
        "clevel": 5,
        "cname": "lz4",
        "id": "blosc",
        "shuffle": 1
    },
    "dimension_separator": "/",
    "dtype": "<u2",
    "fill_value": 0,
    "filters": null,
    "order": "C",
    "shape": [
        3,
        1,
        2160,
        5120
    ],
    "zarr_format": 2
}