ls: Remove potentially redundant syscall, only use direntries for file types

[kimono-koans]

Here, we were requesting symlink_metadata before proceeding to the color_name fn, and I've also completely eliminated the need to hold onto DirEntrys within the PathData struct. Statx calls are faster.

[codspeed-hq]

Merging this PR will degrade performance by 20.22%

⚡ 7 improved benchmarks
❌ 4 regressed benchmarks
✅ 287 untouched benchmarks
⏩ 48 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Memory	ls_recursive_wide_tree[(10000, 1000)]	6.6 MB	5.8 MB	+14.06%
⚡	Memory	ls_recursive_long_all_wide_tree[(15000, 1500)]	9.9 MB	8.7 MB	+14.17%
❌	Memory	ls_recursive_mixed_tree	133.1 KB	163 KB	-18.37%
❌	Memory	ls_recursive_long_all_deep_tree[(100, 4)]	126 KB	158 KB	-20.22%
❌	Memory	ls_recursive_deep_tree[(200, 2)]	143.2 KB	175.1 KB	-18.21%
❌	Memory	ls_recursive_long_all_mixed_tree	132.4 KB	164 KB	-19.25%
⚡	Simulation	ls_recursive_mixed_tree	1.4 ms	1.3 ms	+5.64%
⚡	Simulation	ls_recursive_wide_tree[(10000, 1000)]	53.9 ms	46.5 ms	+15.88%
⚡	Simulation	ls_recursive_deep_tree[(200, 2)]	1.7 ms	1.6 ms	+5.39%
⚡	Simulation	ls_recursive_balanced_tree[(6, 4, 15)]	56.9 ms	50.3 ms	+13.05%
⚡	Simulation	ls_recursive_long_all_wide_tree[(15000, 1500)]	136 ms	121.4 ms	+12%

---

_{Comparing kimono-koans:Redundant-statx-syscalls (b526069) with main (c79e686)}

Footnotes

1. 48 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. ↩