ls: enter_directory should use an iterative loop instead of recursing

[kimono-koans]

Rust doesn't have TCO and this uses more memory and risks a stack overflow. We should be dropping unused PathData when we are done with it.

In another PR, I've reduced the PathData size by more than a kilobyte, but master branch's version is a porker and these allocations are probably slowing down ls.

This is also the first step to concurrent directory search.

[erammos]

Rust doesn't have TCO and this uses more memory and risks a stack overflow. We should be dropping unused PathData when we are done with it.

In another PR, I've reduced the PathData size by more than a kilobyte, but master branch's version is a porker and these allocations are slowing down ls.

It seems that is 1.5x - 2x slower, compared to GNU's ls, even with a small number of nested folders

hyperfine -N --warmup 3 'ls -R uutils_test/nested_5_' 'target/release/ls -R uutils_test/nested_5_'
Benchmark 1: ls -R uutils_test/nested_5_
  Time (mean ± σ):       2.2 ms ±   0.8 ms    [User: 0.3 ms, System: 1.6 ms]
  Range (min … max):     0.9 ms …   5.5 ms    1888 runs

Benchmark 2: target/release/ls -R uutils_test/nested_5_
  Time (mean ± σ):       3.8 ms ±   0.9 ms    [User: 0.8 ms, System: 2.7 ms]
  Range (min … max):     1.5 ms …   5.6 ms    920 runs

Summary
  ls -R uutils_test/nested_5_ ran
    1.72 ± 0.74 times faster than target/release/ls -R uutils_test/nested_5_

[kimono-koans]

It seems that is 1.5x - 2x slower, compared to GNU's ls, even with a small number of nested folders

ls has lots of performance issues. It's a target rich environment. I don't think this recursive fn is a huge issue right now, but in combination with other fixes, it might make an impact.

[kimono-koans]

@erammos Well I was wrong. Much faster on ls run through the coreutils directory. See:

Benchmark 1: ls -alR --color .
  Time (mean ± σ):     586.6 ms ±   5.5 ms    [User: 182.5 ms, System: 403.9 ms]
  Range (min … max):   582.2 ms … 597.5 ms    10 runs

Benchmark 2: ~/program/coreutils/target/release/ls -alR --color .
  Time (mean ± σ):     482.3 ms ±   4.3 ms    [User: 212.7 ms, System: 269.3 ms]
  Range (min … max):   474.3 ms … 487.7 ms    10 runs

  Warning: Ignoring non-zero exit code.

Summary
  ~/program/coreutils/target/release/ls -alR --color . ran
    1.22 ± 0.02 times faster than ls -alR --color .

[oech3]

#11215