Bugfix to define global attributes (and remove some unneeded code)

[SamuelTrahanNOAA]

Bug fix: write global attributes from forecast.

As documented here:

• [rrfs-mpas-jedi] Allow MPASJEDI write analysis into init.nc directly at cold-start cycles NOAA-EMC/rrfs-workflow#949

In the comment chain, it was noted that the current gsl/develop cannot write global attributes.

• [rrfs-mpas-jedi] Allow MPASJEDI write analysis into init.nc directly at cold-start cycles NOAA-EMC/rrfs-workflow#949 (comment)

The bug is caused by a workaround I added to let jedivar write to init.nc. Jedivar was trying to write to undefined global attributes in data mode. My workaround was to not write to attributes unless they're defined. That prevented them from being defined in define mode (since nothing is defined yet).

The fix is to always allow writing global attributes in define mode.

Information on running mandatory regression tests on Jet can be found here and the results pasted below.

regression test case results

PASTE compare_run_testcases AND/OR compare_create_testcases TEXT HERE

[SamuelTrahanNOAA]

@guoqing-noaa - This is the fix for your troubles.

[SamuelTrahanNOAA]

I may have PR'd this to the wrong branch. Regardless, the fix guoqing needs is here.

[SamuelTrahanNOAA]

The branch this needs to be PR'd to isn't in this repository. I'm closing the PR.