perf: improve extraction cache throughput

[Priveetee]

Summary

• cache public extraction endpoints with hashed versioned keys and endpoint-specific TTLs
• add cache hit/miss/store logging for public extraction cache
• coalesce concurrent /streams cache misses per URL to avoid duplicate cold extractions

Verification

• JAVA_HOME=/usr/lib/jvm/java-25-openjdk PATH=/usr/lib/jvm/java-25-openjdk/bin:$PATH ./gradlew clean check shadowJar --no-daemon
• runtime curl validation on /suggestions, /trending, /search, /channel, /comments
• runtime concurrent /streams validation: 8 identical misses completed in one shared extraction window