Bug: `uu-sort` fails to reorder binary lines containing NUL bytes

Summary

• Tool under test: uu-sort 0.3.0
• Impact: When the input stream contains embedded NUL (0x00) and other non-UTF-8 bytes, uu-sort reports success but leaves the data in its original order, whereas GNU sort correctly sorts the lines by their raw byte values.
• Reproducer: Feed the two-line blob below (first line contains 0x00) into both binaries.

# emit the test case with Python so the NUL byte is preserved
python3 - <<'PY' | /usr/bin/sort | hexdump -C
import sys
sys.stdout.buffer.write(bytes([
    0x84, 0x64, 0x00, 0x48, 0x48, 0x48, 0x48, 0x84, 0xf5, 0x0a,
    0xf5, 0xf5, 0xd9, 0x0b, 0xdd, 0xf5, 0xf5, 0xf5, 0xf5,
]))
PY

Expected result (GNU sort)

$ python3 - <<'PY' | /usr/bin/sort | hexdump -C
> import sys
> sys.stdout.buffer.write(bytes([
>     0x84, 0x64, 0x00, 0x48, 0x48, 0x48, 0x48, 0x84, 0xf5, 0x0a,
>     0xf5, 0xf5, 0xd9, 0x0b, 0xdd, 0xf5, 0xf5, 0xf5, 0xf5,
> ]))
> PY
00000000  f5 f5 d9 0b dd f5 f5 f5  f5 0a 84 64 00 48 48 48  |...........d.HHH|
00000010  48 84 f5 0a                                       |H...|
$ echo $?
0

Actual result (uu-sort)

$ python3 - <<'PY' | instrumented-coreutils/target/release/sort | hexdump -C
> import sys
> sys.stdout.buffer.write(bytes([
>     0x84, 0x64, 0x00, 0x48, 0x48, 0x48, 0x48, 0x84, 0xf5, 0x0a,
>     0xf5, 0xf5, 0xd9, 0x0b, 0xdd, 0xf5, 0xf5, 0xf5, 0xf5,
> ]))
> PY
00000000  84 64 00 48 48 48 48 84  f5 0a f5 f5 d9 0b dd f5  |.d.HHHH.........|
00000010  f5 f5 f5 0a                                       |....|
$ echo $?
0

uu-sort echoes the input verbatim (aside from appending a trailing newline) yet still returns success, so scripts that expect sorted output silently receive unsorted data.

Notes

• The issue shows up whenever a line includes a NUL and other invalid UTF-8 bytes: uu-sort likely decodes the data as UTF-8 (or via String::from_utf8_lossy) before comparing lines, causing both lines to collapse to the same comparison key and leaving the original order unchanged.
• Fix idea: compare raw u8 slices rather than UTF-8 Strings so binary inputs remain sortable.

Start at the uu-sort entry point and reproduce the two-line binary input shown in the issue, comparing its output with GNU sort. Trace how input lines become comparison keys; done means embedded NUL and non-UTF-8 bytes are ordered by their raw byte values, with a regression test for the reproducer.