ptx: match GNU's error for a bad --gap-size/--width value

[arbelonson-source]

What

--gap-size/-g and --width/-w validate their values with clap's
own value_parser!(u64).range(1..), which causes the same two
problems already fixed for sort --parallel (#14303) and nl's
numeric options (#14314):

$ ptx -g -5 f

# GNU: still rejects the value, but with its own wording
ptx: invalid gap width: '-5'

# uutils, before this PR: treated as an unrecognized flag instead
error: unexpected argument '-5' found
  tip: to pass '-5' as a value, use '-- -5'

$ ptx -g bogus f

# GNU
ptx: invalid gap width: 'bogus'

# uutils, before this PR
error: invalid value 'bogus' for '--gap-size <NUMBER>': invalid digit found in string

Fix

Resolve the value by hand: parse as u64, reject anything that
doesn't parse as a positive integer (unparseable, zero, negative, or
overflowing) with GNU's wording. Unlike nl --number-width (#14314),
GNU does not distinguish an overflow from any other invalid value
here -- --gap-size 99999999999999999999999999 gets the exact same
"invalid gap width: '...'" message as --gap-size bogus -- so this
one only needed a single uniform check, confirmed by diffing both
cases against real GNU rather than assuming the nl-style distinction
would apply here too.

Testing

• cargo test -p uu_ptx / full tests/by-util/test_ptx.rs suite: 50 passed, 0 failed.
• Strengthened the existing test_invalid_arg test (which only checked the exit code, with a comment noting the message was clap-provided) to assert the exact GNU wording.
• Added a regression test for the separate-argument negative-value case (the actual reported bug) on both -g and --width.
• Manually diffed 12 cases (invalid, negative attached/separate/short-flag, zero, overflow, empty, valid) against GNU ptx 9.11 under LC_ALL=C.
• cargo clippy -p uu_ptx --all-targets -- -D warnings and cargo fmt --check: clean.

---

This PR was written with AI assistance (Claude Opus 5, via Claude Code). I've tested the changes but please review the code carefully.

[arbelonson-source]

Closing proactively, same reasoning as #14314 just now — same template, same maintainer guidance from #14308/#14312.

[github-actions]

GNU testsuite comparison:

Note: The gnu test tests/cp/sparse-2 is now being skipped but was previously passing.
Skip an intermittent issue tests/pr/bounded-memory (was skipped on 'main', now failing)