split: quote a zero SIZE in the invalid-number diagnostic

[Socialpranker]

A SIZE that is rejected for being zero was printed bare, while every other
rejected SIZE is quoted:

$ split --lines 0 file        # GNU
split: invalid number of lines: '0'
$ split --lines 0 file        # uutils, before
split: invalid number of lines: 0

$ split --lines 2fb file      # both, already quoted
split: invalid number of lines: '2fb'

Same for --bytes 0, --line-bytes 0 and the obsolete split -0 spelling.

The fix

parse_size_u64_max quotes the value inside the ParseSizeError it
returns, so the parse-failure path was already right. The zero check
happens after that call and built its own ParseSizeError::ParseFailure
from the raw string, unquoted. It now quotes it the same way. Two call
sites: get_and_parse and the obsolete-lines branch.

Three existing assertions in test_split.rs pinned the unquoted form; they
are updated.

How the GNU behavior was established

By running the installed GNU coreutils 9.11 binary (Homebrew, gsplit) as
a black box over -l/-b/-C/-n with the values 0, abc, 1Y,
-200 and an overflowing number, diffing stderr and exit status against
uutils. I did not read GNU coreutils source.

Testing

• New test_split_zero_size_is_quoted in tests/by-util/test_split.rs
  covers --lines/--bytes/--line-bytes with the exact message and
  exit status 1. Mutation-checked: reverting strategy.rs alone makes it,
  and the two updated tests, fail.
• cargo test --features split --test tests test_split: 135 passed, 0 failed.
• cargo clippy -p uu_split --all-targets -- -D warnings: clean.
• cargo fmt --check: clean.
• Differential A/B against GNU coreutils 9.11 over the 19 split
  invocations in my harness: mismatches 2 -> 0.

Not in scope

split -C 0 reports "invalid number of bytes" where GNU reports "invalid
number of lines", and the obsolete split -0 spelling omits the
Try 'split --help' line GNU prints. Both are separate from the quoting
and left alone here.

Disclosure

Prepared with AI assistance (Claude Opus 5, via Claude Code), per the AI
policy in CONTRIBUTING.md. Every GNU behavior quoted above came from
running the installed binary, not from reading GPL source. All testing was
run locally.

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/cut/bounded-memory (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/date/resolution (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/tail/pipe-f is now passing!