Add JMH benchmark proving direct Math.multiplyHigh path outperforms invokeExact

[peter-lawrey]

This PR adds a JMH benchmark for XXH3 hashing of a 128-byte array and switches the Java 9+ multiplication path from reflective MethodHandle.invokeExact(...) to a direct Math.multiplyHigh(...) call enabled by the Java 8 compile-time stub.

Why

The previous Java 9+ path used reflection plus MethodHandle.invokeExact(...) so the project could still compile for Java 8. That avoided a hard Java 9 API dependency, but it also prevented the hot multiplication path from being as efficiently optimised and inlined as a normal Math.multiplyHigh(...) call.

This change keeps Java 8 compilation support while allowing newer JVMs to call Math.multiplyHigh(...) naturally.

What changed

• Added JMH test dependencies.
• Added XXH3HashBytesBenchmark covering XXH3 hashing of a 128-byte array, reproducing issue Performance degradation between 0.16 and 0.27ea1 #101.
• Replaced the MethodHandle.invokeExact(...) path in MathsJDK9 with direct Math.multiplyHigh(...) calls.
• Retained runtime fallback to the scalar Java 8 implementation when Math.multiplyHigh(...) is unavailable.
• Added SonarCloud project properties.

Expected benefit

On Java 9+ runtimes, XXH3 can now use the JVM-recognised Math.multiplyHigh(...) path directly, making it eligible for normal JIT optimisation/inlining instead of paying the invokeExact(...) overhead in the hashing hot path.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud