Fix resolving reference variables in mixed subgroups

[jasonb5]

Querying a variable whose value is a reference that exists
in a different subgroup would fail with ERROR: Entry HIST_N was listed as type int but value '$STOP_N' is not valid int.

This PR changes the behavior so a variable that cannot
be resolved within the current subgroup will try to resolve
from any subgroup. Finally if still not resolvable an exception
is raised.

Test suite: pytest CIME/tests/test_unit*
Test baseline:
Test namelist changes:
Test status: [bit for bit, roundoff, climate changing]

Fixes #4646

User interface changes?:
Update gh-pages html (Y/N)?:

[jasonb5]

@jedwards4b @jgfouca The original solution discussed for this issue was to "namespace" the value e.g. HIST_N=<subgroup>::STOP_N. I'm wondering if this complexity is required? The change I have right now simply falls back to resolving the reference without a subgroup. Is there a scenario when a variable would be defined more than twice?

[jedwards4b]

I would suggest adding a test for a variable which is in a subgroup, for example
JOB_WALLCLOCK_TIME

 ./xmlquery JOB_WALLCLOCK_TIME
Results in group case.cupid
	JOB_WALLCLOCK_TIME: 48:00:00
Results in group case.st_archive
	JOB_WALLCLOCK_TIME: 00:20:00
Results in group case.test
	JOB_WALLCLOCK_TIME: 48:00:00

[jasonb5]

@jedwards4b The test is there but it's not clear, will fix.

[jasonb5]

Looks like a subgroup reference would be needed for the following example.

Results in group case.cupid
	JOB_WALLCLOCK_TIME: $JOB_WALLCLOCK_TIME@case.cupid
Results in group case.st_archive
	JOB_WALLCLOCK_TIME: 00:20:00
Results in group case.test
	JOB_WALLCLOCK_TIME: 48:00:00

@jedwards4b @jgfouca Any recommendations for the delimiter? Originally it was suggested to use :: like c++, another alternative could be @.

[jgfouca]

@jasonb5 , I think :: would be best. We use that in Eamxx tools already.

[github-actions]

You can preview documentation at https://esmci.github.io/cime/branch/fix_mixed_subgroup_resolve/html/index.html