Can't use templated benchmarking here because the benchmark requires several container instances with different element types.
Code is same between files with only the containers replaced, which can be verified via diff.