Upgrade Hive client to 2.3.9 and HIVE platform JDK toolchain to 17

[YogeshKothari26]

Summary

Bumps the plugin's pinned Hive client 1.2.2 → 2.3.9 and the HIVE platform's JDK toolchain 8 → 17 in Defaults.java. Bytecode for the HIVE platform stays at Java 8 (options.release.set(8)) so the produced consumer UDF jars remain runnable on Java 8 runtimes.

Spark (spark_2.11, spark_2.12) and Trino subprojects are not changed.

Motivation

• Hive 1.2.2 transitively pulls org.pentaho:pentaho-aggdesigner-algorithm:5.1.5-jhyde, which is not resolvable on Maven Central; downstream builds that don't add an explicit exclusion fail to resolve. Hive 2.3.9 is excludable cleanly.
• Hive 1.2.2's embedded HiveServer2 + DataNucleus + Derby reflection paths fail under JDK 17. Hive 2.3.9 is JDK-17-friendly with the standard --add-opens flags applied here.
• Net result: downstream UDF projects can move their build JVM to Java 17 without per-project workaround patches for Hive itself.

Changes

• Defaults.java: HIVE platform JavaLanguageVersion.of(8) → of(17) + add org.pentaho exclusion to the consumer's hive-exec compileOnly configuration.
• TransportPlugin.java: pin bytecode release = 8 for non-Trino Java platforms; add --add-opens JVM args to the test launcher when the platform's JLV is >= 17 and the platform is non-Trino.
• transportable-udfs-plugin/build.gradle: hive-version 1.2.2 → 2.3.9 in the generated version-info.properties.
• transportable-udfs-hive/build.gradle: hive-exec 1.2.2 → 2.3.9 (compileOnly + testImplementation), add org.pentaho exclusion, add --add-opens to the subproject's own test task when the build JVM is JDK 17.
• transportable-udfs-test-hive/build.gradle: hive-exec / hive-service 1.2.2 → 2.3.9, add org.pentaho exclusion.
• HiveTester.java: two Hive 2.x compat fixes — replace the removed FunctionInfo(boolean, String, GenericUDF) ctor with FunctionInfo(FunctionType.PERSISTENT, ...); disable METASTORE_SCHEMA_VERIFICATION and enable datanucleus.schema.autoCreateAll on the embedded HiveConf (Hive 2.3.x's embedded Derby strictly verifies schema version on startup).

Testing

1. End-to-end Spark integration

A Spark cluster job registers Hive UDFs built by this plugin snapshot and runs SQL queries that exercise:

• Scalar in / out — StdString upper-case with null and empty-string edge cases.
• Multi-arg primitives — 2-arg StdLong addition with null-argument propagation.
• Complex output — StdMap<StdString,StdString> construction with key-lookup, size, and null-key cases.

All cases registered via CREATE TEMPORARY FUNCTION + SELECT against a SparkSession. Exercises driver↔executor serialization, Spark's Hive UDF compatibility shim, and metastore-backed function registration end-to-end. All test cases pass.

2. JAR diff vs master

SHA-256 + size + class-set + bytecode-major-version on the 7 shipped plugin artifacts (snapshot vs baseline 0.2.0):

• 5 of 7 plugin jars are CONTENT_IDENTICAL (every class file byte-identical; zip metadata drift only). The 2 jars with deliberate content change are transportable-udfs-plugin.jar (the toolchain wiring) and transportable-udfs-test-hive.jar (the Hive 2.x call-site fixes) — i.e. only the files this PR touches.
• version-info.properties flips hive-version=1.2.2 → 2.3.9 as intended.
• Hive UDF wrapper class file confirmed major version: 52 via javap -v (Java 8 bytecode target — preserves grid runtime compatibility). Trino wrapper unchanged at major version: 61.

3. Plugin example UDFs (build + unit)

./gradlew build test on the PR branch — 51 / 51 example tests pass (17 hiveTest on JDK 17 launcher, 17 trinoTest, 17 generic) across transportable-udfs-examples covering every StdUDF arity + Std type (primitives, StdArray, StdMap, StdStruct, nested). Build is green on both JDK 8 and JDK 17 build JVMs.

Note for consumers bumping their build JVM to JDK 17

The Spark platforms (spark_2.11, spark_2.12) intentionally stay at JavaLanguageVersion.of(8) because Spark 3.1.1 hits SPARK-33772 on JDK 17. If you bump your build JVM to JDK 17 and use either Spark platform, keep a JDK 8 toolchain reachable from Gradle (e.g., add its path to org.gradle.java.installations.paths in your gradle.properties). The HIVE platform itself needs no consumer-side workaround. This pin lifts once the plugin's pinned Spark version is bumped to 3.5.x.