Update support for the GSI on Jet

[DavidHuber-NOAA]

As part of the global workflow support on Jet (NOAA-EMC/global-workflow#357), the Jet build/run scripts and regression tests for the GSI should be updated. The GSI has successfully built and run using the global workflow based on commit 9c1fc15d on port_2_jet. Additional work will be required to incorporate these changes into the master and to support regression tests on Jet, including adding the test case dataset.

@MichaelLueken-NOAA I will add the changes to incorporate the build/run script changes. I'm also willing to make the code changes to enable regression testing on Jet. Do you know if the regression test data is already hosted on Jet?

[MichaelLueken]

@DavidHuber-NOAA I don't have an official allocation on Jet, just my home directory, so there are no regression test canned data available on Jet.

[DavidHuber-NOAA]

I've successfully ported the regression tests and they all pass except global_fv3_4denvar_C192 and global_C96_fv3aero.

For global_fv3_4denvar_C192, it failed because I failed to add add the path to ncdump. This test is rerunning now.

For global_C96_fv3aero, it failed for two reasons:

1. the time it took to run global_C96_fv3aero_hiproc_updat was ~65s, with a threshold of ~39s
2. related, the scalability test failed as the time difference between global_C96_fv3aero_hiproc_updat and global_C96_fv3aero_loproc_updat differed by ~34s, which is greater than the threshold of 4.6s.

@MichaelLueken-NOAA Assuming global_fv3_4denvar_C192 passes, may I proceed with the port or do these threshold violations require more attention?

[MichaelLueken]

@DavidHuber-NOAA So long as the regression tests reproduce, then you should be fine moving forward with the port (wall time and scalability can pass with a resubmission of the tests, failures are generally due to system noise ).

[DavidHuber-NOAA]

The global_fv3_4denvar_C192 test failed on the second run as global_fv3_4denvar_C192_loproc_updat used 3.2GB of hardware memory, exceeding the threshold of 2.6GB. All reproducibility tests passed, so I am going to proceed.

All I have left at this point are the monitors, which I just need to test. I'm going to wrap in the port of the monitors to S4 in this issue as well.