fix: ReverseSuffix whitelist includes CharClass Plus patterns

[kolkov]

Summary

Fix performance regression where [^\s]+\.txt pattern caused extreme benchmark to hang.

Problem

• isSafeForReverseSuffix only recognized .* and .+ wildcards
• CharClass Plus patterns like [^\s]+ were not recognized as safe wildcards
• Result: Pattern fell back to slow path (266ms/MB instead of µs)

Solution

• Add CharClass Plus patterns to whitelist of safe wildcard patterns
• Now patterns like [^\s]+\.txt, [\w]+\.go use ReverseSuffix optimization

Testing

• Added TestReverseSuffix_CharClassPlus test
• Local benchmark: suffix_find now completes in 398µs (was timing out)
• All existing tests pass

Impact

• Fixes extreme benchmark CI hang
• No regressions in other patterns

[github-actions]

Benchmark Comparison

Comparing main → PR #94

Summary: geomean 252.6n 251.9n -0.30%

⚠️ Potential regressions detected:

geomean                               ³                +0.00%               ³
geomean                               ³                +0.00%               ³
geomean                         ³                +0.00%               ³
geomean                         ³                +0.00%               ³
AnchoredAlt_ManyBranches_Stdlib/NoMatch-4               85.95n ± ∞ ¹    86.51n ± ∞ ¹     +0.65% (p=0.032 n=5)
FatTeddyFallback/large_haystack_1KB-4                   308.8n ± ∞ ¹    312.6n ± ∞ ¹     +1.23% (p=0.008 n=5)
IPRegex_Find/coregex_64KB_sparse-4                      3.107µ ± ∞ ¹    3.358µ ± ∞ ¹     +8.08% (p=0.008 n=5)
IPRegex_Find/coregex_1MB_sparse-4                       2.733µ ± ∞ ¹    3.040µ ± ∞ ¹    +11.23% (p=0.008 n=5)
IPRegex_Find/stdlib_6MB_sparse-4                        128.0µ ± ∞ ¹   3506.6µ ± ∞ ¹  +2640.06% (p=0.008 n=5)
IPRegex_Find/coregex_6MB_sparse-4                       3.638µ ± ∞ ¹    6.134µ ± ∞ ¹    +68.61% (p=0.008 n=5)

Full results available in workflow artifacts. CI runners have ~10-20% variance.
For accurate benchmarks, run locally: ./scripts/bench.sh --compare