cat: improve splice fast-path

[oech3]

Fixes #11516

$ yes-gf421d0112 | cat-p | pv >/dev/null
^C.4GiB 0:00:04 [23.1GiB/s]
$ yes-gf421d0112 | cat | pv >/dev/null
^C.6GiB 0:00:05 [6.33GiB/s]

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/cut/bounded-memory (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/date/date-locale-hour (passes in this run but fails in the 'main' branch)
Note: The gnu test tests/basenc/bounded-memory is now being skipped but was previously passing.

[codspeed-hq]

Merging this PR will improve performance by 13.59%

⚠️ 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

⚡ 1 improved benchmark
✅ 299 untouched benchmarks
⏩ 46 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Memory	cp_recursive_deep_tree[(120, 4)]	699.2 KB	615.6 KB	+13.59%

---

_{Comparing oech3:splice-pipe (e4331a4) with main (af6c8f2)²}

Footnotes

1. 46 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. ↩

2. No successful run was found on main (2b05f2f) during the generation of this report, so af6c8f2 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[github-actions]

GNU testsuite comparison:

Note: The gnu test tests/pr/bounded-memory is now being skipped but was previously passing.
Congrats! The gnu test tests/tail/pipe-f is now passing!

[oech3]

truncate -s 9EB huge; cat-p huge | pv>/dev/null is 3~10GiB/s faster than zero-copy yes now...

$ taskset -c 1 cat-p huge |taskset -c 2 pv>/dev/null
^C39GiB 0:00:06 [40.0GiB/s]
$ taskset -c 1 cat huge |taskset -c 2 pv>/dev/null
^C92GiB 0:00:15 [20.7GiB/s]

[oech3]

I have no idea how to achieve yes which has same perf with cat sparse-huge.
I considered that tee()ing 1 MiB pipe with references of y is same with splice()ing a zero page, but it has simllar perf with real 1 MiB y.