remove hardcoded compclasses

[jedwards4b]

Removes some lingering hardcoded lists of COMP_CLASSES and uses get_values instead
Test suite: scripts_regression_tests.py
Test baseline:
Test namelist changes:
Test status: bit for bit

Fixes #1170
Fixes #1171

User interface changes?:

Code review:

[gold2718]

Did you test all of these modified scripts? I don't think scripts_regression_tests.py tests them all (e.g., ncr).

[jedwards4b]

Testing of ncr indicates that it didn't work before this PR. Would you like me to jump down this rabbit hole?

[gold2718]

Well, it sort of says that our testing program is not very robust. At the very least, you should document the status in this PR.
How about the other tests you changed? Were they all tested? Please document that.

[jedwards4b]

Rabbit hole jumping ... tomorrow.

[gold2718]

Why was this merged before the requests were fulfilled? This PR still is not correctly documented and a new issue should be opened for the failing test.

[fischer-ncar]

I merged it so I could run the cesm2_0_alpha06g overnight.

[gold2718]

Why was this PR required? We always seem to be rushing around and now we have code in master which either has not been tested or at the least has tests which are undocumented.
What's the rush?

[fischer-ncar]

I messed up. I thought Jim told be to go ahead and make a tag with this PR. It was really without
this PR.

[fischer-ncar]

I reverted this PR, but I'm can't reopen it. I think Jim needs to make a new PR. Sorry about
all of this.

[rljacob]

Also you'll need to "revert the revert" when you finally put this branch on master.

[fischer-ncar]

Which is better, to "revert the revert", or make a new PR?

[rljacob]

If the PR is for the same branch, you'll still have to revert the revert. See https://git-scm.com/blog/2010/03/02/undoing-merges.html