Satellite phenology mode run fails with gnu compiler

[glemieux]

Additional testing of satellite phenology mode with E3SM using the gnu compiler resulted in the follow issue: E3SM-Project/E3SM#4755. SP mode hadn't been tested with the gnu compiler as both our clm and elm regression tests for this mode utilize the intel compiler.

Replicating the issue on Cori the results in the following error immediately during initialization:

378:  in sub-canopy layer in leafc_from_treelai
378:  this is not set up to work for lower canopy layers.
378:  problem in leafc_from_treelai         379           2
378:  ENDRUN:ERROR in /global/u1/g/glemieux/E3SM/components/elm/src/external_models/fates/biogeochem/FatesAllometryMod.F90 at line 811                                                                                                                                                                                                                                                                                                                                                       

Tracing the code, I determined that the issue has to do with the cohort%canopy_layer being un-initialized and not specific to elm-fates. This is pretty clear from the error message above as the 379 is the canopy_layer trying to be indexed here, which doesn't make sense for fates generally. I confirmed this with a test run on Cheyenne as well. I've implemented a simple fix to assign the temporary starting canopy_layer value to be one, but this has exposed a couple other issues with sp mode initialization as well that need to be addressed.

Note: sp mode will run with the gnu compiler if the case or test is set up in debug mode.

[rosiealice]

nice job @glemieux :)