Update UDF compiler for shared helper layout

[gerashegalov]

Related to #14834.

Description

This PR is one reviewable layer in the unshim stack introduced by #15025. It updates the UDF compiler and related documentation for the shared helper layout. This is the final follow-up layer on top of the unshim stack.

Stack context

• Stack introduction: Add default common unshim packaging flow #15025
• Previous layer: Adapt Delta and Iceberg to unshimmed helpers #15027

Testing and validation notes

• Covered by the full-stack packaging/build validation described in Add default common unshim packaging flow #15025 and the existing UDF compiler coverage for the affected integration points.
• 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 adapts the UDF compiler module to the "shared helper layout" introduced by the unshim stack. The main changes remove the dependency on Spark's internal Logging trait (replacing it with direct SLF4J calls via LoggerFactory), switch Cast and Encode from companion-object apply to direct new construction with explicit None timezone, convert case class LogicalPlanRules() to class LogicalPlanRules (compatible with reflection-based instantiation), and update documentation links.

• Logging migration: All UDF compiler classes (BB, Instruction, CatalystExpressionBuilder, GpuScalaUDFLogical, LogicalPlanRules) drop with Logging and instead declare a static SLF4J logger in a companion object; isDebugEnabled/isTraceEnabled guards are correctly added since Scala string interpolation is eager unlike Spark's by-name logDebug.
• Cast/Encode construction: Companion-object apply calls (e.g. Cast(e, t)) are replaced with new Cast(e, t, None) and new Encode(...), reflecting a layout change that no longer exposes the Spark companion objects directly; timezone is None in all cases where it was previously also the default.
• Doc updates: Pandas UDF links updated from the archived Spark 3.2.0 docs to live Spark 3.5.7 docs, and the AQE code snippet is corrected to use new GpuQueryStagePrepOverrides (matching the case class → class pattern in the prior stack layer).

Confidence Score: 5/5

This PR is safe to merge — all changes are mechanical refactors with no logic mutations.

The changes swap Spark's internal Logging trait for direct SLF4J calls, convert Cast/Encode companion-object apply to new construction (preserving all defaults), and convert one case class to a plain class that is only ever instantiated via reflection. All call sites have been verified, isDebugEnabled guards are correctly added, and the timeZoneId forwarding in CatalystExpressionBuilder.simplifyExpr is preserved unchanged.

No files require special attention; both Scala 2.12 and 2.13 Instruction.scala variants are in sync.

Important Files Changed

Filename	Overview
udf-compiler/src/main/scala/com/nvidia/spark/udf/LogicalPlanRules.scala	Changed from case class LogicalPlanRules() to class LogicalPlanRules; instantiated via reflection in ShimLoader so no direct call-site impact; Logging trait removed cleanly.
udf-compiler/src/main/scala/com/nvidia/spark/udf/GpuScalaUDF.scala	Logging moved to companion object logger; logDebug guarded with isDebugEnabled, logWarning replaced with direct warn (no guard needed at warn level).
udf-compiler/src/main/scala/com/nvidia/spark/udf/CatalystExpressionBuilder.scala	Logging and Cast/Encode construction updated for shared helper layout; timezone is correctly forwarded (ce.timeZoneId) where the original used it, and None elsewhere.
udf-compiler/src/main/scala/com/nvidia/spark/udf/CFG.scala	Added object BB companion with SLF4J logger; all trace/debug calls wrapped with level guards; no logic changes.
udf-compiler/src/main/scala-2.12/com/nvidia/spark/udf/Instruction.scala	Scala 2.12 copy: same Logging→SLF4J and Cast/Encode constructor changes as the shared file; in sync with 2.13 variant.
udf-compiler/src/main/scala-2.13/com/nvidia/spark/udf/Instruction.scala	Scala 2.13 copy: identical changes to 2.12 variant; both kept in sync.
udf-compiler/src/main/scala/com/nvidia/spark/udf/State.scala	Copyright year update and two Cast(l, t) calls updated to new Cast(l, t, None); semantics preserved.
docs/additional-functionality/rapids-udfs.md	All Pandas UDF documentation links updated from archive.apache.org/dist/spark/docs/3.2.0 to spark.apache.org/docs/3.5.7; no content changes.
docs/dev/adaptive-query.md	Code snippet updated from GpuQueryStagePrepOverrides() companion-object style to new GpuQueryStagePrepOverrides, matching the prior stack layer's class change.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[UDF Compiler Module] --> B[LogicalPlanRules\nclass - no-arg ctor]
    B --> C[GpuScalaUDFLogical\ncase class + companion log]
    C --> D[CatalystExpressionBuilder\ncase class + companion log]
    D --> E[CFG / BB\ncase class + companion log]
    D --> F[Instruction\ncase class + companion log]
    D --> G[State]

    subgraph Logging
        H[Before: extends Logging\nSpark internal trait]
        I[After: LoggerFactory.getLogger\nSLF4J direct - shared helper]
    end

    subgraph CastConstruction
        J["Before: Cast(e, t)\ncompanion apply"]
        K["After: new Cast(e, t, None)\ndirect constructor"]
    end

    B -.->|reflection| L[ShimLoader.newInstanceOf]
    H --> I
    J --> K

Loading

_{Reviews (1): Last reviewed commit: "Update UDF compiler for shared helper la..." | Re-trigger Greptile}