Adapt remaining SQL plugin callers to Java helpers

[gerashegalov]

Related to #14834.

Description

This PR is one reviewable layer in the unshim stack introduced by #15025. It updates the remaining SQL plugin callers to use the moved Java helpers. This closes out the caller migration before the common shim-helper cleanup begins.

Stack context

• Stack introduction: Add default common unshim packaging flow #15025
• Previous layer: Adapt SQL RAPIDS callers to columnar helpers #15033
• Next layer: Move common shim helpers to shared sources #15052

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 is one layer of the unshim stack: it migrates the remaining SQL plugin callers off Spark's private Logging mixin (replaced with direct slf4j.LoggerFactory or the project's own RapidsLocalLog) and off Scala case class definitions that blocked cross-module Java interop (replaced with plain class + explicit val fields and Serializable).

• All ProfileMsg subclasses, LoreRDD*, AggAndReplace, BoundGpuWindowFunction, ParsedBoundary, GpuUnboundedToUnboundedAggStages, and the window exec helper classes are migrated from case class to class; pattern-match sites updated to use class-based matching with field accessors.
• DateTimeUtilsShims.currentTimestamp is inlined as a local Instant.now() computation, and the hasAliasQuoteFix shim branch in GpuAlias.sql is removed since every supported Spark version returns true for that flag.
• ParquetCachedBatchSerializer adds the Buffer cast for ByteBuffer.position() to restore Java 8 / 9+ cross-version compatibility.

Confidence Score: 4/5

• Safe to review in the stack context; the only actionable concern is that GpuLiteralShim is referenced but not yet defined, so this layer cannot compile standalone.
• The bulk of the change is a mechanical, well-scoped migration — logging traits, case-class-to-class conversions, and pattern-match sites. All examined call sites are consistent with the new APIs. The one concrete problem is that GpuLiteral now extends GpuLiteralShim, a trait imported from com.nvidia.spark.rapids.shims that has no definition anywhere in the current HEAD (no Scala file, no Java file, no shim version directory). Any build that compiles this layer independently will fail. Because this is an intentional forward reference to the next stack layer (Move common shim helpers to shared sources #15052), it only blocks merging this PR before that one lands.
• sql-plugin/src/main/scala/com/nvidia/spark/rapids/literals.scala — the unresolved GpuLiteralShim import needs attention before this layer can be built in isolation.

Important Files Changed

Filename	Overview
sql-plugin/src/main/scala/com/nvidia/spark/rapids/literals.scala	GpuLiteral changed to extend GpuLiteralShim (removing jsonFields override), but GpuLiteralShim is not defined anywhere in the current HEAD — forward dependency on next stack layer.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/profiler.scala	ProfileMsg case classes removed (moved to Java in sql-plugin-columnar); Scala callers updated to use class-based pattern matching and Java method access; logging migrated from Spark Logging mixin to slf4j directly.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/window/BasicWindowCalc.scala	AggAndReplace, BoundGpuWindowFunction, ParsedBoundary migrated from case class to class with explicit equals/hashCode; behaviour is consistent with prior case-class semantics for array fields.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/DateUtils.scala	currentTimestamp shim replaced with a local Instant.now()-based implementation; result semantics (UTC microseconds since epoch) match the shim's behaviour.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/namedExpressions.scala	Removed hasAliasQuoteFix shim branch from GpuAlias.sql; safe because every supported Spark version (3.2.0+) returns true for that flag.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/sparkRapidsListeners.scala	SparkRapidsBuildInfoEvent and SparkRapidsShuffleDiskSavingsEvent migrated from case class to class; the default parameters for the disk-savings event were removed — the single existing call site in ShuffleCleanupManager already passes all arguments explicitly.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/parquet/ParquetCachedBatchSerializer.scala	ByteBuffer.position() calls cast to Buffer for Java 9+ covariant-return compatibility; CloseableColumnBatchIterator migrated from case class to class.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/NvtxWithMetrics.scala	MetricRange primary constructor default for excludeMetric replaced by an auxiliary constructor; all existing call sites verified to be compatible.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/lore/GpuLore.scala	LoreRDDMeta and LoreRDDPartitionMeta migrated from case class to class with Serializable; logging migrated from Spark Logging to slf4j; all object constructions updated correctly.
sql-plugin/src/main/scala/com/nvidia/spark/rapids/window/GpuWindowExpression.scala	BatchedRunningWindowFixer subclasses migrated from Spark Logging to RapidsLocalLog; AggAndReplace construction updated to use new keyword.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    subgraph Before["Before (Spark private Logging + case class)"]
        A["extends Logging\n(Spark internal)"] --> B["case class ProfileInitMsg\ncase class LoreRDDMeta\ncase class AggAndReplace\ncase class BoundGpuWindowFunction\n..."]
        B --> C["Pattern match via\nunapply / case class deconstruct"]
    end

    subgraph After["After (slf4j + Java-friendly class)"]
        D["RapidsLocalLog trait\nor inline slf4j.LoggerFactory"] --> E["class ProfileInitMsg (Java)\nclass LoreRDDMeta\nclass AggAndReplace\nclass BoundGpuWindowFunction\n..."]
        E --> F["Pattern match via\ncase msg: Foo =>  msg.field"]
    end

    subgraph Missing["Forward dependency (HEAD)"]
        G["GpuLiteral extends GpuLiteralShim"] -.->|"GpuLiteralShim not yet defined\n(expected in PR #15052)"| H["⚠ Compilation gap"]
    end

    Before --> After

Loading

_{Reviews (1): Last reviewed commit: "Update Spark 4.1 generated support metad..." | Re-trigger Greptile}

[greptile-apps]

Forward reference to undefined GpuLiteralShim

GpuLiteralShim is imported from com.nvidia.spark.rapids.shims and GpuLiteral now extends it, but a search across the entire repository (all Scala and Java sources, all shim version directories) finds no definition of this trait anywhere in the current HEAD. Building this layer in isolation — or any branch that cherry-picks or rebases onto it before PR #15052 lands — will fail with an unresolved symbol error. If the definition is intentionally deferred to the "common shim-helper cleanup" next layer, a // TODO defined in #15052 comment or a temporary stub would make the dependency explicit.

[greptile-apps]

Plain arithmetic instead of overflow-safe multiply

Spark's own instantToMicros uses Math.multiplyExact and Math.addExact to guard against overflow. The replacement here uses plain * and +. For timestamps within a practical range the result is identical, but it silently returns a wrong (wrapped) value rather than throwing if a pathological timestamp is ever passed, while the shim it replaces would have thrown. Worth noting since this replaces a shim that was previously calling through to Spark's overflow-checked path.