[BUG] loading SPARK-32639 example parquet file triggers a JVM crash

[gerashegalov]

Describe the bug
Replaying the scenario from #1463 (SPARK-32639) in GPU-enabled Scala Spark or pyspark shell results in SIGSEGV in the executor call path

C  [cudf_io5390148209696334138.so+0x170830]  cudf::io::detail::parquet::reader::impl::decode_page_data(hostdevice_vector<cudf::io::parquet::gpu::ColumnChunkDesc>&, hostdevice_vector<cudf::io::parquet::gpu::PageInfo>&, hostdevice_vec
tor<cudf::io::parquet::gpu::PageNestingInfo>&, unsigned long, unsigned long, rmm::cuda_stream_view)+0x4b0

Java frames: (J=compiled Java code, j=interpreted, Vv=VM code)
j  ai.rapids.cudf.Table.readParquet([Ljava/lang/String;Ljava/lang/String;JJIZ)[J+0
j  ai.rapids.cudf.Table.readParquet(Lai/rapids/cudf/ParquetOptions;Lai/rapids/cudf/HostMemoryBuffer;JJ)Lai/rapids/cudf/Table;+122
j  com.nvidia.spark.rapids.MultiFileParquetPartitionReader.$anonfun$readToTable$1(Lai/rapids/cudf/ParquetOptions;Lai/rapids/cudf/HostMemoryBuffer;JLcom/nvidia/spark/rapids/NvtxWithMetrics;)Lai/rapids/cudf/Table;+4
j  com.nvidia.spark.rapids.MultiFileParquetPartitionReader$$Lambda$2380.apply(Ljava/lang/Object;)Ljava/lang/Object;+16
j  com.nvidia.spark.rapids.Arm.withResource(Ljava/lang/AutoCloseable;Lscala/Function1;)Ljava/lang/Object;+2

Steps/Code to reproduce bug

1. Start (py)Spark REPL with RAPIDS plugin and GPU enabled

2. load the file from attached to SPARK-32639

spark.read.schema('value MAP<STRUCT<first:STRING, middle:STRING, last:STRING>, STRING>').parquet("/home/gshegalov/gits/spark-rapids/integration_tests/src/test/resources/SPARK-32639/000.snappy.parquet").take(1)

Expected behavior
In Spark rc 3.1.1 the data loads fine on CPU and should do the same on GPU:

>>> spark.read.schema('value MAP<STRUCT<first:STRING, middle:STRING, last:STRING>, STRING>').parquet("/home/gshegalov/gits/spark-rapids/integration_tests/src/test/resources/SPARK-32639/000.snappy.parquet").take(1)
[Row(value={Row(first='John', middle='Y.', last='Doe'): 'brother'})]

Environment details (please complete the following information)

• Environment location: Local master/shell
• Spark configuration settings related to the issue:

	--conf spark.plugins=com.nvidia.spark.SQLPlugin \
	--conf spark.rapids.sql.enabled=true \
	--jars  ${SPARK_CUDF_JAR},${SPARK_RAPIDS_PLUGIN_JAR}

[jlowe]

cudf fixed the crash in rapidsai/cudf#7229. We just need to re-enable the test when the cudf jar picks up that fix.