Append NWM Forecast subset to Parquet and Zarr

[echeipesh]

Wanted script to extract one or more HUCs from NWM short-range prediction NetCDF format and append to existing .parquet dataset.

I think the hard part of this issue is to figure out if/how it's possible to append to Parquet file from python and what the schema for the streams file thats friendly to appending should be.

Assumptions:

• clip_nwm.py will at some point be called in AWS Batch or similar with required parameters to extract multiple HUCs
• Because NetCDF file has only one prediction window, it makes sense to read it once and extract multiple HUCs instead of calling it once per HUC
• We want to append to Parquet file at least on RowGroup Level

Questions:

• What is the layout of the Parquet file?
• Is it actually possible to append to Parquet from python libraries?
• Should we be trying to append at all?
  • How big is a Parquet file for a Delaware basin? (Maybe its large enough that we don't want to grow it)

Notes: It is not clear its easy to append to Parquet files. Lots of SO examples talk about re-reading and re-writing the file. That's not an option because we expect to read It appears possible based on this Java implementation: https://github.com/apache/parquet-mr/pull/278 It does not appear possible to do this using PyArrow

[lewfish]

We can append to Parquet and Zarr. See linked PRs.