fix: FindAll/Count O(n²) for $-anchored patterns

[kolkov]

Summary

FindAll/Count for $-anchored patterns were O(n²) — each iteration ran PikeVM from successive positions to end of input.

Root cause: UseReverseAnchored was missing from findIndicesAtWithState switch, falling through to NFA. Match/Find were correct (single reverse scan via findIndicesReverseAnchored).

Fix: short-circuit in findAllIndicesLoop and Count — a $-anchored pattern without (?m) matches at most once. One reverse searcher call, at most one result, no loop.

ERROR.*$ on 2400-line log: 3.3 s → instant.

Test plan

• 5 correctness cases vs stdlib (match, no-match, case-insensitive, single-line)
• Multiple-candidate test: 3 "ERROR" occurrences, stdlib returns 1 match at end
• Linear-time verification on 2400 lines (would be 3.3 s if quadratic)
• Full suite green, lint clean
• regex-bench CI: 4/4 green

Fixes #183

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

Benchmark Comparison

Comparing main → PR #185

Summary: geomean 104.6n 105.1n +0.46%

⚠️ Potential regressions detected:

geomean                    104.6n         105.1n        +0.46%
geomean                               ³                +0.00%               ³
geomean                               ³                +0.00%               ³
geomean                         ³                +0.00%               ³
geomean                         ³                +0.00%               ³
AnchoredLiteralVsStdlib/coregex_long-4                  9.131n ± ∞ ¹    9.343n ± ∞ ¹    +2.32% (p=0.008 n=5)
ASCIIOptimization_Issue79/medium_WithASCII-4            630.6n ± ∞ ¹    644.2n ± ∞ ¹    +2.16% (p=0.008 n=5)
LangArenaLogParser/post_requests-4                      1.181m ± ∞ ¹    1.183m ± ∞ ¹    +0.19% (p=0.016 n=5)
LangArenaLogParser/auth_attempts-4                      808.1µ ± ∞ ¹    812.5µ ± ∞ ¹    +0.55% (p=0.032 n=5)
LangArenaLogParser/sessions-4                           482.1µ ± ∞ ¹    486.8µ ± ∞ ¹    +0.96% (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