ls: match GNU's error for a bad --format/--sort/--time/etc. value

[arbelonson-source]

What

Eight options in ls validate an enumerated choice with a
ShortcutValueParser: --format, --hyperlink, --quoting-style,
--time, --sort, --color, --indicator-style, and --classify.
An unrecognized or ambiguous value on any of them produces clap's own
generic wording instead of GNU's grouped-alias one, e.g.:

$ ls --time=bogus

# GNU
ls: invalid argument 'bogus' for '--time'
Valid arguments are:
  - 'atime', 'access', 'use'
  - 'ctime', 'status'
  - 'mtime', 'modification'
  - 'birth', 'creation'
Try 'ls --help' for more information.

# uutils, before this PR
error: invalid value 'bogus' for '--time <field>'

  [possible values: atime, ctime, mtime, birth]

For more information, try '--help'.

Fix

Unlike the other options fixed in this series (#14293, #14303-#14312),
none of these eight are resolved by hand: each keeps its
ShortcutValueParser completely unchanged, since that's still the
source of truth clap uses both to validate the value and to render
e.g. --quoting-style's own --help text (several of these already
carry real synonym handling clap needs to keep doing correctly).

Instead, clap errors now go through a small wrapper (get_matches)
that, on an ErrorKind::InvalidValue naming one of these eight
options, reconstructs GNU's wording from a small metadata table
(CHOICE_OPTIONS) describing each option's own choice groups in GNU's
display order. Every other clap error -- unknown flags, --help,
--version, etc. -- still goes through the exact same shared
formatter as before, unchanged. This mirrors the approach already
used for tee --output-error (#14308), which has the same
help-text constraint.

Two things specific to these eight, found only by diffing exhaustively
against real GNU rather than just checking one bad value per option:

• --format=columns is this implementation's own choice; GNU's real
  --format has no separate word for what it calls vertical and
  does not accept columns at all. Listed as its own group, same
  precedent as du's creation/birth (du: match GNU's error for a bad --time value, fix --time-style's help hint #14306) -- and, as with that
  one, it creates a real ambiguity GNU doesn't have (e.g.
  --format=co resolves to commas on GNU, but is ambiguous here
  between commas and columns).
• --quoting-style is unusual in the other direction: GNU lists
  each of its choices, including c and c-maybe -- a real alias
  pair -- on its own separate line, rather than grouping aliases
  together the way every other option in this series does.

Testing

• cargo test -p uu_ls / full tests/by-util/test_ls.rs suite: 189 passed, 0 failed, 1 pre-existing ignore.
• Updated the existing test_localized_possible_values test (which asserted clap's own wording, localized, as the expected behavior) to assert GNU's wording instead, in both English and French.
• Added 7 new regression tests covering each of the three message shapes (flat, grouped, one choice per line) and confirming --classify=WHEN parsing itself still works.
• Manually diffed every one of the 8 options across invalid, ambiguous, empty, and valid-abbreviation values against GNU ls 9.11 under LC_ALL=C (including cases specific to each option's own alias structure, e.g. --time=a unambiguous vs --format=v ambiguous).
• Manually confirmed ordinary usage (-la, --color=auto, --time-style=...), --help, -h, --version, and an unrelated error (unknown flag, still exit code 2) are all unaffected.
• cargo clippy -p uu_ls --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.

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/retry (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/cut/bounded-memory (passes in this run but fails in the 'main' branch)
Skip an intermittent issue tests/pr/bounded-memory (was skipped on 'main', now failing)

[oech3]

We don't need to show all equivalent aliases on help. Such big code is really no worth to maintain. Please close this PR.