fix(C++): Fix reversed parameters and incorrect calls in string_util benchmark

[lzaeh]

What does this PR do?

This PR fixes several minor issues and improves the implementation of string utility benchmarks:

• Corrected reversed function parameters.
• Fixed an incorrect function call.
• Replaced some usages of auto with explicit types for better readability and maintainability.
• Refactored random number generation by encapsulating std::mt19937 in a thread-local helper function.

One part of the original code used default random generation methods, which could result in uneven value distributions and non-deterministic behavior across platforms. To improve randomness quality and thread-safety, I replaced it with a thread_local encapsulated std::mt19937 generator. This ensures better reproducibility and consistency during benchmarking.

After these changes, the code compiles and runs correctly. Benchmark results are stable and as expected. (See image below.)

Related issues

• [C++] Improve and enhance the fury-cpp benchmark module #2245

Does this PR introduce any user-facing change?

• Does this PR introduce any public API change?
• Does this PR introduce any binary protocol compatibility change?

Benchmark

[pandalee99]

LGTM