cmake: replace -fopenmp-target-fast with -O3 for LLVMFlang AMD GPU builds

[sbryngelson]

Summary

Replaces -fopenmp-target-fast with -O3 in the LLVMFlang OpenMP target offload compile flags.

-fopenmp-target-fast is present in the minimal reproducer for the array constructor corruption bug reported in #1449. Replacing it with explicit -O3 preserves optimization level while avoiding the flag combination that triggers silent wrong values in GPU_PARALLEL_LOOP regions on AMD MI250X (gfx90a).

Closes #1449

Test plan

• Build with ./mfc.sh build --gpu mp -j 8 on Frontier using AMD flang
• Run IBM test cases (./mfc.sh test --only IBM -j 8) and confirm ghost point corrections are correct
• Run full test suite on Frontier to check for regressions

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 61.31%. Comparing base (8e34431) to head (84a1bff).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1450   +/-   ##
=======================================
  Coverage   61.31%   61.31%           
=======================================
  Files          72       72           
  Lines       19771    19771           
  Branches     2852     2852           
=======================================
  Hits        12123    12123           
  Misses       5699     5699           
  Partials     1949     1949           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.