[WIP]: Initial prototype

[TomAugspurger]

This is a prototype of what a stactools package like this might look like.

At a high level, it would make one STAC item per NetCDF file distributed by NODD. For example, "https://noaanwm.blob.core.windows.net/nwm/nwm.20231010/short_range/nwm.t00z.short_range.channel_rt.f001.conus.nc".

It relies on that naming convention to derive some STAC properties, using the parts after /nwm/, i.e. nwm.20231010/short_range/nwm.t00z.short_range.channel_rt.f001.conus.nc.

I've optionally used a new feature of xstac to embed Kerchunk references in the STAC metadata. See examples/example-item/example-item.json for an example.

Some open questions:

1. I've invented some names for "product" (channel_rt, land, etc.), "Region" (consus, alaska, etc), "Category" ("short_range", "long_range", etc.), "cycle runtime". These probably have proper names given by NOAA. We should use those.
2. What's the appropriate bbox and geometry to use for the channel_rt data? We can't get it just from the NetCDF data. We'd need to get the reaches file. It should be the same for every item (by region: conus, alaska, etc.).
3. We should use the forecast extension to catalog the reference_datetime, horizon, etc.
4. I've assumed that we'd have one STAC Catalog per "category - product" pair. But I haven't implemented that, and we could be flexible with how this is modeled.