Calculating toa incident solar radiation as well as updating gdas_utility products unit

[SadeghTabas-NOAA]

Description

This PR updates gdas_utility.py to calculate TOA incident solar radiation based on datetime, lat, and lon using the PySolar package and puts it in the graphcast inputs. This process is parallelized, as it may take time for high resolution and longer steps.

Also, this PR updates gdas_utility.py to generate graphcast gdas inputs with the same units as ERA5 that the graphcast is trained on. Unit updated variables:

• geopotential_at_surface (m2/s2)
• geopotential_at_levels (m2/s2)
• total_precipitation_6hr (m)

In addition, some updates have been done to resolve the memory issue due to merging a large number of nc files when we generate gdas info for 10 days (40 steps).

Linked Issues and Pull Requests

Associated GraphCast Issue to close

• Closes https://github.com/NOAA-EMC/graphcast/issues/11

Blocking Dependencies

Anticipated Changes

Input data

• [ ] No changes are expected to input data.
• [x] Changes are expected to input data:
  • [ ] New input data.
  • [x] Updated input data.

Needed libraries

• pysolar
• joblib

[junwang-noaa]

Also you mentioned using GFS for normalization, will it be a separate PR?

[SadeghTabas-NOAA]

Also you mentioned using GFS for normalization, will it be a separate PR?

I'd update normalization part with the script for fine tuning if its possible?

[junwang-noaa]

The code looks good, thanks!