Fix benchmark runner formatting check

[wlmrh]

Summary

• format benchmark/llama_engine_benchmark.cpp with clang-format 14.0.6
• address the formatting failure observed in PR Realistic benchmark reporting #6 and the merged main CI run
• make no runtime, benchmark behavior, or report-data changes

Verification

• full repository clang-format-14 --dry-run --Werror check passes
• Python syntax check passes
• 21 benchmark utility tests pass
• CPU warnings-as-errors build completes locally

Local platform note

The macOS CTest run completed 59 tests successfully, skipped 7 model-dependent tests, and retained one pre-existing platform-sensitive RoPE tolerance failure (1.24e-5 difference against a 1e-5 tolerance). This PR does not modify that test or runtime behavior.