Github Actions New CI Comparison Method

[scrasmussen]

WARNING: PR HAS CHANGED A LOT, SEE COMMENTS LATER IN PR FOR CURRENT STATUS

TYPE: bug fix

KEYWORDS: CI, Google Drive, Croton Testcase

SOURCE: Soren Rasmussen, NCAR

DESCRIPTION OF CHANGES: Switch to gdown python package to download croton testcase. This new method is easier and better suppported.

TESTS CONDUCTED: Tested new lines in a test Python script and they succesfully download the Croton testcase.

NOTES: I believe the Docker image will need to be changed to make the Python package gdown a requirement so it downloads it. Will add that after submitting this pull request.

[scrasmussen]

During debugging PR has changed a lot from original purpose

DESCRIPTION OF CHANGES:

• Github Actions no longer use Docker to package dependencies and run tests. It does the installation of dependencies itself with apt-get and pip. It handles the building and running of reference and candidate code, comparison of outputs.
• Parts no longer used by CI: Docker, wrfhydropy, the following files: test/local/run_tests.py, test/test_*.py. Reduces the runtime by around half
• tests/compare_output.py has the Python scripts for comparing output. This replaces the nccmp functionality
• The Croton, NY testcase that is part of WRF-Hydro releases is downloaded from Github. Google Drive is no longer needed and the functionality in Python to get the files off of the drive can be removed
• CMake adds Makefile targets during the build process to build the code and run with the Croton testcase. Targets follow the $ make croton-gridded, $ make run-croton-gridded and $ make run-croton-gridded-parallel format. This is done for gridded, gridded_no_lakes, nwm, nwm_ana, nwm_long_range, reach, reach_lakes

TESTS CONDUCTED: Current build is comparing reference to reference output since the CI will fail on the candidate code

NOTES: checklist for acceptance and testing of this PR

• debugging cleanup
• add back in Run Reference Command Github Action step
• merge PR
• create new test PR with small change to produce different output. This will test the reporting capacity of this PR (sometimes the Python scripts fail but it doesn't register that failure in the Github Action step)

[scrasmussen]

Reference has been added back in, three github action reports to note

• Comparing every outputted file takes over 11 minutes
• Comparing the last file takes slightly over 3 minutes
• Adding back in comparison to reference comparison, rather than comparison of candidate to candidate, breaks as expected