JMH based benchmarks

Use JMH support in MX to run our JMH benchmarks.

New JMH benchmarks: context initialization and host interop micro benchmarks.

Update the existing JMH interop benchmarks to not be single shot and let JMH
harness run the benchmark loop instead of the benchmark function itself. Remove
the custom JMH driver and rely on the standard JMH entry point (used by JMH MX
support). Replace explicit Blackhole usage with returning a value, which should
be equivalent. Keep only PyEuler11 for now, other benchmarks seem to exercise
the same/similar interop interactions.

Author: steve-s

graalpython/com.oracle.graal.python.benchmarks/java/com/oracle/graal/python/benchmarks/JavaBenchmarkDriver.java

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021, 2021, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2021, 2024, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * The Universal Permissive License (UPL), Version 1.0
@@ -58,6 +58,11 @@
 import org.graalvm.polyglot.Source;
 import org.graalvm.polyglot.Value;
 
+/**
+ * Driver that can execute our Python benchmarks via Java embedding. Java version of the Python
+ * benchmark harness. The benchmarks must define Python function entry point named
+ * {@code java_embedded_bench_entrypoint} and this driver executes that as one benchmark iteration.
+ */
 public class JavaBenchmarkDriver extends AbstractLanguageLauncher {
     private static final String LINE = "-------------------------------------------------------------------------------";
     private static final String BENCHMARK_ENTRY_POINT = "java_embedded_bench_entrypoint";