nl: output number separator as bytes

[cakebaker]

Currently, we output the number separator as a lossy string:

$ printf "a\nb" | cargo run -q nl --number-separator=$'\xFF' | hexdump -C
00000000  20 20 20 20 20 31 ef bf  bd 61 0a 20 20 20 20 20  |     1...a.     |
00000010  32 ef bf bd 62 0a                                 |2...b.|
00000016

Whereas GNU nl doesn't:

$ printf "a\nb" | nl --number-separator=$'\xFF' | hexdump -C
00000000  20 20 20 20 20 31 ff 61  0a 20 20 20 20 20 32 ff  |     1.a.     2.|
00000010  62 0a                                             |b.|
00000012

This PR fixes the issue.

Update: The PR also fixes the same issue for the content. Our current output:

$ printf $'\xFF' | cargo run -q nl | hexdump -C
00000000  20 20 20 20 20 31 09 ef  bf bd 0a                 |     1.....|
0000000b 

The output of GNU nl:

$ printf $'\xFF' | nl | hexdump -C
00000000  20 20 20 20 20 31 09 ff  0a                       |     1...|
00000009

[codspeed-hq]

CodSpeed Performance Report

Merging #9601 will improve performances by 66.23%

_{Comparing cakebaker:nl_fix_non_utf8_number_separator (db52eac) with main (aaa0610)}

Summary

⚡ 2 improvements
✅ 125 untouched
⏩ 6 skipped¹

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	nl_large_file[10]	100.9 ms	61.2 ms	+65.03%
⚡	nl_many_lines[100000]	79.8 ms	48 ms	+66.23%

Footnotes

1. 6 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:

Skipping an intermittent issue tests/tail/overlay-headers (passes in this run but fails in the 'main' branch)

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/overlay-headers (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/cksum/b2sum is no longer failing!

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/overlay-headers (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/cksum/b2sum is no longer failing!

[ChrisDryden]

I was hoping we could create a macro that is similar to the new_ucmd that allows us to validate using the args and and then make sure that the stderr, stdout and error code match with the GNU implementation, I believe that would have highlighted this issue earlier right?

[ChrisDryden]

Amazing performance improvement, we should look for more changes like this in the codebase. For anyone looking in the future for more perf improvements its now at the point that a big percentage of the time is spent freeing the vec, and if you do .clear it actually just sets the len to 0 so that you can just use the same buffer in the loop and add more capacity if needed by checking buf.capacity().