Remove dead ACC loop directives in s_derive_center_of_mass

[sbryngelson]

Summary

• In s_derive_center_of_mass (m_derived_variables.fpp), two $:GPU_LOOP(parallelism='[seq]') directives appeared before the c_m zero-initialization loops, outside any GPU parallel region.
• These expanded to !$acc loop seq with no surrounding !$acc parallel/!$acc kernels, so the compiler silently ignored them. CCE 19.0.0 correctly warns: ftn-7255: loop marked for accelerator worksharing but not inside an accelerator region.
• The initialization runs on CPU and is pushed to device via the existing $:GPU_UPDATE(device='[c_m]') immediately after — the structure is correct, the directives were just dead code.
• Replace the 7-line nested loop with c_m(:, :) = 0.0_wp, which is idiomatic Fortran and eliminates the spurious directives.

Test plan

• Verify ftn-7255 warning is gone from the CCE 19.0.0 (cpe/25.03) build of simulation
• Confirm s_derive_center_of_mass produces identical results (center-of-mass values unchanged)

🤖 Generated with Claude Code

[github-actions]

Claude Code Review

Head SHA: b1dc5c9

Files changed: 1

• src/simulation/m_derived_variables.fpp

Summary:

• Removes 7 lines of dead GPU loop directives and replaces them with 1 idiomatic Fortran array assignment
• The original code used $:GPU_LOOP(parallelism='[seq]') directives outside any GPU parallel region — they expanded to dangling ! loop seq without a surrounding ! parallel/! kernels, making them no-ops that triggered a CCE 19.0.0 compiler warning (ftn-7255)
• The replacement c_m(:, :) = 0.0_wp is correct Fortran, removes the spurious directives, and leaves the existing $:GPU_UPDATE(device='[c_m]') downstream untouched — the semantics are unchanged
• Net change: -6 lines, cleaner code, silences a legitimate compiler warning

Findings:

No issues found. The change is correct and minimal.

Improvement opportunities (optional):

• None — this is as clean as it gets for this fix.