v11: Add better exit control for gcm_run.j

[mathomp4]

Closes #774

This is an attempt to try to have better exit control from gcm_run.j

---

Closes #818

In re #818, this PR tries to update all the "bad" exit calls in gcm_run.j to have a non-zero status. Hopefully this leads to better trapping with cylc, et al.

cc: @Dooruk

[Dooruk]

I'm testing with this, my code failed gracefully this time.

4222              18  18  18  18  18  18
4223 NOTE from PE     0: MOM_restart: MOM run restarted using : INPUT/MOM.res_3.nc
4224
4225 FATAL from PE     0: MOM_restart: Unable to find mandatory variable age in restart files.
4226
4227 GEOSgcm.x failed with return code 1

[Dooruk]

Welp, I'm testing with v12 but no reason for this not to work for v11?

[mathomp4]

Welp, I'm testing with v12 but no reason for this not to work for v11?

Yeah, it's already in v12 (#776) but now that I know it works, I've undrafted this and hopefully @sdrabenh will pull it in in the next update.

Thanks for confirming it!

[Dooruk]

I spoke to soon, this is Cylc not capturing the error if nodes are not set properly. Cylc doesn't register that as a fail (says succeed but actually it stopped running). Should this be a separate issue (again, i'm testing with v12rc**)?:

[mathomp4]

I spoke to soon, this is Cylc not capturing the error if nodes are not set properly. Cylc doesn't register that as a fail (says succeed but actually it stopped running). Should this be a separate issue (again, i'm testing with v12rc**)?:

Ugh. yeah. Make an issue. I can try and trace how that's not working.