test: add BMI2 bitwise benchmark baseline

[tzh476]

This is the baseline PR requested in https://github.com/QuEST-Kit/QuEST/pull/779#issuecomment-2948379234.\n\nIt adds the same automated benchmark from #779, but is based on current devel and intentionally does not include the BMI2 helper implementation changes. The goal is to let CI run the identical benchmark against the current fallback/original bitwise helper paths for a like-for-like comparison with #779.\n\nValidation performed locally:\n- Confirmed the benchmark source matches #779 byte-for-byte.\n- Confirmed the CMake patch matches the benchmark-only commit from #779 via stable patch-id.\n- Ran git diff --check.\n\nI could not run the CMake build locally because this machine does not currently have cmake installed; CI should provide the cross-platform benchmark output.

[tzh476]

The baseline workflows are currently in action_required state (audit, compile, and test-free). Could you approve/run them when convenient so we can collect the fallback benchmark numbers for comparison with #779? Thanks!

[tzh476]

For visibility: the upstream workflows for this baseline PR still appear to need maintainer approval, but I ran the same benchmark from this branch in a temporary fork-only workflow to get provisional fallback numbers:

• https://github.com/tzh476/QuEST/actions/runs/27348202274/attempts/3

That run used an AMD EPYC 7763 runner with Ubuntu 24.04 / GCC 13.3.0 and printed BMI2 intrinsics: disabled, as expected for the baseline/no-helper branch. I posted the provisional same-CPU comparison against #779 here: #779 (comment)

The authoritative signal should still come from this PR's upstream CI once approved.