grep: avoid multi-character case folds

[wondr-wclabs]

Fixes #32.

This changes the ignore-case compile path to use Oniguruma's per-regex onig_new_deluxe API with the default case-fold flags minus INTERNAL_ONIGENC_CASE_FOLD_MULTI_CHAR.

The important distinction is that this is not only an output formatting issue. With Oniguruma's default full-fold behavior, one regex atom can consume multiple input characters: [[:alpha:]] can match st as a single match, and ß can match SS. That changes -o output, match counts, and cursor advancement. GNU grep avoids those multi-character fold expansions for grep -i, so the matcher needs to stay on simple one-to-one folds even before broader locale support exists.

I avoided ONIG_OPTION_IGNORECASE_IS_ASCII because that would be a broader behavioral change: it would discard non-ASCII simple folds too, while the incompatibility here is specifically the multi-character fold expansion. I also avoided onig_set_default_case_fold_flag because it is process-global; using compile-time OnigCompileInfo keeps the change local to each compiled regex and avoids races between different pattern compilations.

The small raw wrapper mirrors the existing onig::Regex shape where relevant: it serializes regex construction with a mutex, keeps the same Send/Sync ownership model for immutable compiled regexes, reuses onig::Region's transparent raw representation, and formats errors through Oniguruma's own error-message API.

Validation:

• cargo fmt --all -- --check
• cargo test
• cargo clippy --all-targets --workspace -puu_grep -- -D warnings
• git diff --check
• printf 'st\nss\nffi\n' | cargo run --quiet -- -o -i '[[:alpha:]]' now emits one character per line
• printf 'SS\n' | cargo run --quiet -- -o -i 'ß' exits 1 with no output

[codspeed-hq]

Merging this PR will improve performance by 22.08%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 6 improved benchmarks
✅ 4 untouched benchmarks
⏩ 17 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	recursive_no_binary	29.3 ms	22.3 ms	+31.07%
⚡	regex_no_match	29.6 ms	22.7 ms	+30.59%
⚡	context	32.1 ms	25.1 ms	+27.58%
⚡	only_matching	54.9 ms	46.8 ms	+17.35%
⚡	filename_lineno_color	55 ms	46.9 ms	+17.32%
⚡	invert_match	75 ms	68.1 ms	+10.11%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing wondr-wclabs:codex/simple-case-folding (7bae963) with main (35b6dc6)}

Footnotes

1. 17 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. ↩

[Abusalah0]

suggestion: Could we add a control test for a non-ASCII one-to-one case fold, such as ä matching Ä with -i?

The current tests verify that multi-character folds like ß → SS are disabled, but they do not verify the other side of the intended behavior: that ordinary Unicode case-insensitive matching is still preserved.

[sylvestre]

@copilot resolve the merge conflicts in this pull request