shuf: prepare args outside benchmark

[xtqqczze]

Split from #10388

[oech3]

Thankyou!

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/tail/retry is no longer failing!

[ChrisDryden]

When I'm reading the code changes the memory benefit is intuitively there, but when I'm looking at the comparisons I don't see any difference. Is the expectation that this will only impact a small subset of the benchmarks? https://codspeed.io/uutils/coreutils/runs/compare/6977c5ceca5d5cfb0f1ce7a5..69779602f6180a419cc41917?uri=src%2Fuu%2Fshuf%2Fbenches%2Fshuf_bench.rs%3A%3Ashuf_lines%255B100000%255D&runnerMode=Memory&q=mode%3Amemory+shuf

[xtqqczze]

We shouldn’t be allocating unnecessarily to prepare arguments inside the measured section. We don’t want to be measuring the performance of the allocator itself, so any allocation work should happen before the benchmark runs. In this case it’s quite possible we aren’t allocating enough to see a meaningful difference.

[sylvestre]

i am not convinced why we should land this PR, could you please explain a bit more ? :)

[oech3]

Separated to investigate #10486