Add Spark 3.3.0 SQL shim module sources

[gerashegalov]

Related to #14834.

Description

This PR is one reviewable layer in the unshim stack introduced by #15025. It adds the Spark 3.3.0 SQL shim module sources after the shared helper module wiring is in place. Keeping this as its own layer makes the Spark-family source population easy to inspect.

Stack context

• Stack introduction: Add default common unshim packaging flow #15025
• Previous layer: Move Hadoop file I/O helpers to Java module #15030
• Next layer: Add Spark 3.3.1 through 3.4 SQL shim module sources #15041

Testing and validation notes

• No standalone behavior change is intended in this layer. It is covered by the full-stack packaging/build validation described in Add default common unshim packaging flow #15025 and the existing tests for the affected subsystem.
• The full split stack was verified to be tree-equivalent to the pre-split stack top.

Checklists

Documentation

• Updated for new or modified user-facing features or behaviors
• No user-facing change

Testing

• Added or modified tests to cover new code paths
• Covered by existing tests
  (Covered by the validation notes in the PR description.)
• Not required

Performance

• Tests ran and results are added in the PR description
• Issue filed with a link in the PR description
• Not required

[greptile-apps]

Greptile Summary

This PR populates the sql-plugin-shims module with Spark 3.3.0 and Databricks 3.3.0 (11.3) SQL shim sources as one reviewable layer in the unshim stack. All files are pure additions with no deletions.

• Spark 3.3.0 shims: SparkShimServiceProvider, ShuffleManagerShims, ShuffleClientShims, FileCommitProtocolShims, OriginContextShim, SparkUpgradeExceptionShims, SequenceSizeTooLongErrorBuilder, CreateDataSourceTableAsSelectRules, and TrampolineConnectShims are added; most carry broad version headers covering 330–411.
• Databricks 3.3.0 shims: DatabricksShimServiceProvider, SparkShimServiceProvider (330db), OriginContextShim, SparkDateTimeExceptionShims, and SparkUpgradeExceptionShims (330db-specific constructor) are added.
• Shared helpers: ConvUtils (reflection-based overflowInConvError dispatcher) and SparkSessionUtils (reflection-based session/leafNodeDefaultParallelism accessor) land in the non-version-gated source tree.

Confidence Score: 3/5

The bulk of the shim wiring is safe; two files have defects that could surface silently at runtime.

SparkDateTimeExceptionShims on Databricks 3.3 silently discards messageParameters, producing exceptions with empty substitution values. SparkSessionUtils.invokeNoArg leaves NoSuchMethodException unhandled so callers targeting Spark 3.3 see a raw propagated exception with no fallback opportunity.

SparkDateTimeExceptionShims.scala (spark330db) and SparkSessionUtils.scala (shared)

Important Files Changed

Filename	Overview
sql-plugin-shims/src/main/spark330db/scala/org/apache/spark/sql/rapids/shims/SparkDateTimeExceptionShims.scala	DB-specific SparkDateTimeException factory; the messageParameters argument is silently dropped, producing exceptions with blank substitution values.
sql-plugin-shims/src/main/scala/org/apache/spark/sql/rapids/shims/SparkSessionUtils.scala	Reflection-based helper; only catches InvocationTargetException, leaving NoSuchMethodException unhandled for Spark 3.4+-only methods like leafNodeDefaultParallelism.
sql-plugin-shims/src/main/scala/org/apache/spark/sql/errors/ConvUtils.scala	Shared reflection shim for overflowInConvError; passes null for single parameter without type validation, risking NPE on primitive-boxed parameter types.
sql-plugin-shims/src/main/spark330/scala/org/apache/spark/sql/rapids/shims/TrampolineConnectShims.scala	SparkSession/Dataset bridge for pre-Connect Spark; type aliases shadow local imports with identical types, minor redundancy only.
sql-plugin-shims/src/main/spark330db/scala/com/nvidia/spark/rapids/DatabricksShimServiceProvider.scala	Databricks runtime detection via BuildInfo; correctly uses ignoreExceptions guard.

Comments Outside Diff (1)

1. sql-plugin-shims/src/main/scala/org/apache/spark/sql/rapids/shims/SparkSessionUtils.scala, line 86-93 (link)

   NoSuchMethodException not caught in invokeNoArg

   Class.getMethod(name) throws NoSuchMethodException when the method is absent, which this helper does not catch. leafNodeDefaultParallelism was introduced in Spark 3.4; calling it against a Spark 3.3 runtime will propagate a raw NoSuchMethodException with no fallback. Catching NoSuchMethodException | IllegalAccessException and wrapping in UnsupportedOperationException would make version-gated call sites fail more informatively.

_{Reviews (1): Last reviewed commit: "Add SQL shim module sources for Spark 3...." | Re-trigger Greptile}

[greptile-apps]

messageParameters silently dropped

The messageParameters: Map[String, String] argument is accepted by the method but never forwarded to the SparkDateTimeException constructor — Array.empty is passed in its place. Any caller that supplies message parameters expecting them to appear in the exception message will silently get an exception with empty/blank substitution values, producing a misleading or incomplete error string at runtime on Databricks 3.3.0.

[greptile-apps]

Hardcoded null passed to unknown parameter type

The code finds any overflowInConvError overload with exactly one parameter and invokes it with null. If that parameter is a primitive-boxed type, JVM unboxing will throw a NullPointerException inside invoke, masking the intended error. Either matching the exact parameter type in the find predicate or adding a comment documenting the nullable assumption would reduce the ambiguity.