Various fixes and improvements to the save-zarr command

[eschombu]

This PR substantially reworks the save-zarr command so that a chunked volume can be saved in a zarr store and successfully loaded and viewed in neuroglancer. Here is an example for viewing the image data at s3://ccn-connectomics-mega-ng/sample3_funke_lab/07_yuri_v5_B.zarr:
https://neuroglancer-demo.appspot.com/#!%7B%22dimensions%22:%7B%22x%22:%5B8e-9%2C%22m%22%5D%2C%22y%22:%5B8e-9%2C%22m%22%5D%2C%22z%22:%5B8e-9%2C%22m%22%5D%7D%2C%22position%22:%5B6875.0%2C4500.0%2C5630.0%5D%2C%22layers%22:%5B%7B%22type%22:%22image%22%2C%22source%22:%22zarr://s3://ccn-connectomics-mega-ng/sample3_funke_lab/07_yuri_v5_B.zarr/%22%2C%22tab%22:%22source%22%2C%22name%22:%2207_yuri_v5_B%22%2C%22shader%22:%22%23uicontrol%20invlerp%20normalized%5Cnvoid%20main%28%29%20%7B%5Cn%20%20emitGrayscale%28normalized%28%29%29%3B%5Cn%7D%5Cn%22%7D%5D%2C%22layout%22:%224panel%22%7D

The following features and fixes were included:

• zyx -> xyz ordering (modifiable using an order argument)
• setting the voxel scale (in nm) as an attribute in the .zattrs
• setting the dtype
• inferring the dtype from the input chunk
• writing chunks in parallel (for use with slurm/disBatch)