bench: fix variance in remaining numfmt benchmarks

[sylvestre]

No description provided.

[codspeed-hq]

CodSpeed Performance Report

Merging this PR will degrade performance by 4.92%

_{Comparing sylvestre:numfmt (2b20379) with main (cd288fc)¹}

Summary

⚡ 2 improved benchmarks
❌ 6 regressed benchmarks
✅ 274 untouched benchmarks
⏩ 38 skipped benchmarks²

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Memory	numfmt_to_si_precision[10000]	4.6 MB	4.8 MB	-3.2%
❌	Memory	numfmt_round_modes[("up", 10000)]	4.8 MB	4.9 MB	-3.16%
❌	Memory	numfmt_to_si[10000]	4.7 MB	5 MB	-4.92%
⚡	Memory	cp_large_file[16]	119 KB	110 KB	+8.23%
❌	Simulation	numfmt_to_iec[10000]	34 ms	35.1 ms	-3.21%
❌	Simulation	numfmt_to_si[10000]	34 ms	35.1 ms	-3.24%
❌	Simulation	numfmt_from_si[10000]	32.6 ms	33.7 ms	-3.27%
⚡	Memory	dd_copy_default	130.8 KB	126 KB	+3.78%

Footnotes

1. No successful run was found on main (f2bec7b) during the generation of this report, so cd288fc was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

2. 38 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩