Fix reductions ctest for nvidia on perlmutter

[amametjanov]

Release build with -O2 flag (from E3SM) incorrectly optimized out inner-loop if-true-stmt.
This re-orders for-loops to prevent branch prediction.

Fixes #161

Checklist

• Testing
  • A comment in the PR documents testing used to verify the changes including any tests that are added/modified/impacted.
  • Unit tests have passed. Please provide a relevant CDash build entry for verification.

[amametjanov]

Testing:

• nvidia on pm-cpu:
  • release- and debug-builds:

25/29 Test #25: REDUCTIONS_TEST ....................   Passed    3.19 sec
...
25/29 Test #25: REDUCTIONS_TEST ....................   Passed   39.95 sec

• nvidiagpu on pm-gpu:
  • release- and debug-builds:

25/29 Test #25: REDUCTIONS_TEST ....................   Passed  108.85 sec
...
25/29 Test #25: REDUCTIONS_TEST ....................   Passed   62.27 sec

[mark-petersen]

Thanks! I was able to repeat the tests above and they indeed work.