segfaults when trying to use PIO with mpiuni with the cray (cce) compiler with ESMF_BOPT=O

[billsacks]

Following the merge of #205 , there are new unit test failures with mpiuni with the cray (cce) compiler on derecho with ESMF_BOPT=O (tests with ESMF_BOPT=g pass).

The new failures are these:

CRASHED: mpiuni/O: src/Infrastructure/Field/tests/ESMC_FieldGridRegridCsrvUTest.c
CRASHED: mpiuni/O: src/Infrastructure/Field/tests/ESMC_FieldRegridUTest.c
CRASHED: mpiuni/O: src/Infrastructure/Field/tests/ESMC_FieldSMMFromFileUTest.c
CRASHED: mpiuni/O: src/Infrastructure/Field/tests/ESMF_FieldSMMFromFileUTest.F90
CRASHED: mpiuni/O: src/Infrastructure/IO/tests/ESMCI_IO_PIOUTest.C
CRASHED: mpiuni/O: src/Infrastructure/IO/tests/ESMF_IOUTest.F90
CRASHED: mpiuni/O: src/Infrastructure/Mesh/tests/ESMCI_MeshUTest.C
CRASHED: mpiuni/O: src/Infrastructure/Mesh/tests/ESMF_MeshFileIOUTest.F90

CRASHED: mpiuni/O: src/Superstructure/PreESMFMod/tests/ESMF_FileRegridUTest.F90
CRASHED: mpiuni/O: src/Superstructure/PreESMFMod/tests/ESMF_RegridWeightGenUTest.F90
CRASHED: mpiuni/O: src/addon/NUOPC/tests/ESMF_NUOPC_UTest.F90

I have checked all of them, and they all fail at the point when they invoke a routine that would go into PIO. Prior to #205 this would have been skipped.

This may not be worth fixing – especially since these pass with ESMF_BOPT=g, so tracking this down could be challenging, and it could be a compiler bug – but I'm opening this to document the issue.