tee: add splice() and tee() zero-copy fast-path

[oech3]

Closes #11466 .

> tee </tmp/huge |pv >/dev/null
^C58GiB 0:00:01 [2.58GiB/s] 
> tee-splice </tmp/huge |pv >/dev/null
^C.6GiB 0:00:02 [38.4GiB/s]

[codspeed-hq]

Merging this PR will improve performance by 19.71%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 2 improved benchmarks
✅ 95 untouched benchmarks
⏩ 266 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	ls_recursive_balanced_tree[(6, 4, 15)]	50.9 ms	48.9 ms	+4.01%
⚡	Simulation	tee_stdin_file[10000000]	232.8 µs	169 µs	+37.77%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing oech3:tee-splice-m (179a204) with main (d09f351)}

Footnotes

1. 266 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:

GNU test failed: tests/tail/pid-pipe. tests/tail/pid-pipe is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/pr/bounded-memory (was skipped on 'main', now failing)

[oech3]

I'm considering something like

curl-splice url://to/lage.iso | tee /dev/nvmeXnYm /dev/nvmeZnWm > saved.iso

if you feel this is a joke. Unfortunately, I don't know how to get curl-splice.

[oech3]

Simulation tee_stdin_file[10000000] 232.8 µs 169 µs +37.77%

CodSpeed is not a good tool for measuring I/O performance. I think pv or time is much better for the case.