Use Zarrita for v3 data generation

[maxrjones]

@freeman-lab The data/chunking for the new stores from zarrita should be the same as in this branch with a few exceptions:

• The string data formats aren't supported
• The Zarr Python inner/outer chunks specification for 3d.chunked.mixed.compressed.sharded.i2.zarr (shown below) didn't make sense given our last conversation. Please let me know if you'd like a different chunking scheme relative to the zarrita version also shown below.

store = DirectoryStoreV3("data/v3/3d.chunked.mixed.compressed.sharded.i2.zarr")
sharding_transformer = ShardingStorageTransformer("indexed", chunks_per_shard=(3, 3, 1))
z = zarr.array(
    arange(27).reshape(3, 3, 3),
    dtype="i2",
    store=store,
    chunks=(3, 3, 1),
    compressor=Zlib(),
    storage_transformers=[sharding_transformer],
)

data = arange(27, dtype="i2").reshape(3, 3, 3)
path = "3d.chunked.mixed.compressed.sharded.i2.zarr"
a = asyncio.run(
    zarrita.Array.create_async(
        store / path,
        shape=data.shape,
        dtype=data.dtype,
        chunk_shape=(3, 3, 3),
        codecs=[
            zarrita.codecs.sharding_codec(
                chunk_shape=(3, 3, 1),
                codecs=[
                    zarrita.codecs.blosc_codec(
                        typesize=data.dtype.itemsize, cname="zlib"
                    )
                ],
            )
        ],
    )
)
a[:, :, :] = data