[feature] Performance optimizations and benchmark

[StellaLupus]

• KotlinEngine: pre-compile regex patterns once at init instead of per scan (~46% faster)
• Matchers: move per-call regex/list constructions to object-level constants
• CardBins: List -> Set for O(1) BIN lookup
• HyperScanEngine: drop unnecessary toMutableList() conversion
• Add BenchmarkTest covering scan throughput, init speed and match consistency
• CI: post benchmark results as PR comment via actions/github-script

[github-actions]

Benchmark Results

Scan Throughput

Corpus: testText.txt × 250 ≈ 354 KB | Matchers: 65 | Warmup: 5 | Iterations: 20

Engine	mean	p50	σ	Throughput	Matches/iter
KotlinEngine	2955.0 ms	2948.2	20.9	0.12 MB/s	10500
HyperScanEngine	140.8 ms	140.5	3.7	2.46 MB/s	10500
HyperScan(preload)	140.0 ms	140.9	2.7	2.47 MB/s	10500

Init Speed

Matchers: 65 | Warmup: 1 | Iterations: 5

Engine	mean	p50	min	max	σ
KotlinEngine	6.5 ms	5.5	4.7	9.9	1.9
HyperScanEngine	17975.9 ms	18004.0	17782.1	18139.1	140.6
HyperScan(preload)	104.2 ms	103.8	101.9	107.9	2.0

Pre-compiled DB speedup: 172.6x

Match Consistency (testText.txt)

Engine	Matches
KotlinEngine	42
HyperScanEngine	42
HyperScan(preload)	42

equal | preload matches compile

[github-actions]

✅ SNAPSHOT published.

Published version: 1.5.0-SNAPSHOT

Gradle:

dependencies { implementation("org.angryscan:core:1.5.0-SNAPSHOT") }

Maven:

<dependency>
  <groupId>org.angryscan</groupId>
  <artifactId>core</artifactId>
  <version>1.5.0-SNAPSHOT</version>
</dependency>

Commit: 2a9a4dfb21fa71d18a1db7f51787a6120fce4cd7