Add inner iteration loops to amortize CodSpeed instrumentation overhead

Sub-microsecond benchmarks are dominated by the ~60ns per-iteration
valgrind start/stop cost in CodSpeed simulation mode. Add kInnerIterations
(1000) inner loops to all fast benchmarks so the actual work dominates.
Move DoNotOptimize calls outside inner loops to prevent artificial overhead.

Also address review feedback:
- Use tokenless CodSpeed (public repo, no CODSPEED_TOKEN needed)
- Fix warning message to show component-specific path
- Fix stray ". :" in error message
- Verify pinned SHA on re-runs to prevent stale checkouts

.github/workflows/ci.yml

@@ -340,7 +340,6 @@ jobs:
         uses: CodSpeedHQ/action@281164b0f014a4e7badd2c02cecad9b595b70537 # v4
         with:
           run: ${{ steps.build.outputs.binary }}
-          token: ${{ secrets.CODSPEED_TOKEN }}
           mode: simulation
 
   clang-tidy-single: