[SkipRecovery] Recover repeated-key JSON extractor tests [databricks]

[wjxiz1992]

JaCoCo sql-plugin line coverage: +2 lines (GpuJsonTuple.scala; focused repeated-key json_tuple IT, shim 350)

Contributes to NVIDIA/cudf-spark#11361.

Depends on NVIDIA/cudf-spark-jni#4915. This PR remains draft until the JNI change is merged and available to the build.

Summary

• Request LAST_NON_NULL named-field matching from the JNI multi-path API used by GpuJsonTuple.
• Recover the repeated-key cases in:
  • test_get_json_object_formats
  • test_get_json_object_child_formats
  • test_json_tuple_formats
• Keep the remaining 18 issue-11361 exemptions for from_json and JSON scan paths. This is a partial recovery, not closure of the root issue.

Validation

Validation used the exact JNI artifact built from the dependency PR (sha256 cc47495dd18e769d96242738bcf8e3320290b53ec8f87163b42b93c5a514359a).

• Offline dist,integration_tests Maven package, shim 350: BUILD SUCCESS (11/11 reactor modules, 03:05)
• Three recovered cases: 3 passed in 4.94s
• Broader repeated_columns cluster: 12 passed, 31 xfailed, 992 deselected, 6 warnings in 20.97s
• Focused JaCoCo run: 1 passed in 4.44s
• Added production lines covered by the focused run: 2 of 4
• python3 -m py_compile and git diff --check: passed

Performance impact

The Spark-side change adds one enum argument to the existing batched JNI call. It does not add JNI/native calls, input or output columns, allocations, or copies. LAST_NON_NULL scans duplicate named fields to match Spark semantics; no throughput benchmark was run.

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
  (Please provide the names of the existing tests 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