cat: avoid unnecessary alloc on Linux

[oech3]

Split a code path which is rarely called to a different function instead of using vec! like #11675 .

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/tail/retry. tests/tail/retry is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tail/tail-n0f (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/tty/tty-eof (passes in this run but fails in the 'main' branch)

[xtqqczze]

We should probably double check CodSpeed benchmarks with hyperfine for anything allocation related.

[oech3]

cat_default[10000000]
src/uu/cat/benches/cat_bench.rs
0%
89.4 µs
89.3 µs
cat_default[10000]
src/uu/cat/benches/cat_bench.rs
0%
197.3 µs
197.1 µs

[oech3]

Check unnecessary alloc bypass:

$ target/release/cat8192 /etc/hostname
m
$ strace -o /dev/null -e inject=splice:error=ENOSYS target/release/cat8192 /etc/hostname
fish: Job 1, 'strace -o /dev/null -e inject=s…' terminated by signal SIGSEGV (Address boundary error)

[oech3]

> truncate -s 0 /tmp/huge
> taskset -c 1 hyperfine -N "target/release/cat-main /tmp/huge" 'target/release/cat64 /tmp/huge'
Benchmark 1: target/release/cat-main /tmp/huge
  Time (mean ± σ):       1.0 ms ±   0.2 ms    [User: 0.4 ms, System: 0.5 ms]
  Range (min … max):     0.8 ms …   3.6 ms    2736 runs
 
Benchmark 2: target/release/cat64 /tmp/huge
  Time (mean ± σ):     999.4 µs ± 207.1 µs    [User: 382.1 µs, System: 494.9 µs]
  Range (min … max):   819.8 µs … 3966.6 µs    1844 runs
 
Summary
  target/release/cat64 /tmp/huge ran
    1.01 ± 0.30 times faster than target/release/cat-main /tmp/huge