bench: reduce memory variance in cp and numfmt benchmarks

[sylvestre]

No description provided.

[codspeed-hq]

Merging this PR will degrade performance by 8.24%

⚡ 1 improved benchmark
❌ 3 regressed benchmarks
✅ 278 untouched benchmarks
⏩ 38 skipped benchmarks¹

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	cp_large_file[16]	365.6 µs	383.4 µs	-4.64%
❌	Memory	numfmt_large_numbers_si[10000]	4.6 MB	5 MB	-8.24%
⚡	Memory	numfmt_to_si_precision[10000]	4.8 MB	4.6 MB	+4.34%
❌	Simulation	numfmt_large_numbers_si[10000]	35.1 ms	37.8 ms	-7.01%

---

_{Comparing sylvestre:bench4 (2cb20e0) with main (2a044db)}

Footnotes

1. 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. ↩

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tty/tty-eof (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/tail/retry is no longer failing!