[FEA] SortArray supports nested types

[wbo4958]

Could we have the feature for supporting SortArray on nested types

repro

      val data = Seq(
        Row(Row("Adam ", "", "Green"), 1, "M", 10),
        Row(Row("Bob ", "Middle", "Green"), 2, "M", 20),
        Row(Row("Cathy ", "", "Green"), 3, "F", 30)
      )

      val schema = (new StructType()
        .add("name", new StructType()
          .add("firstname", StringType)
          .add("middlename", StringType)
          .add("lastname", StringType))
        .add("low", IntegerType)
        .add("gender", StringType)
        .add("high", IntegerType))

      val df = spark.createDataFrame(spark.sparkContext.parallelize(data), schema)
      df.write.format("parquet").mode("overwrite").save("/tmp/testparquet")

      val df2 = spark.read.parquet("/tmp/testparquet")
      df2.createOrReplaceTempView("df2")
      val df3 = spark.sql("SELECT sort_array(collect_set(struct(low,high))) FROM df2 group by gender")
      df3.explain(true)

explain log

!Exec <ObjectHashAggregateExec> cannot run on GPU because not all expressions can be replaced
  @Expression <AttributeReference> gender#18 could run on GPU
  @Expression <AttributeReference> gender#18 could run on GPU
  @Expression <AggregateExpression> collect_set(struct(low, low#17, high, high#19), 0, 0) could run on GPU
    @Expression <CollectSet> collect_set(struct(low, low#17, high, high#19), 0, 0) could run on GPU
      @Expression <CreateNamedStruct> struct(low, low#17, high, high#19) could run on GPU
        @Expression <Literal> low could run on GPU
        @Expression <AttributeReference> low#17 could run on GPU
        @Expression <Literal> high could run on GPU
        @Expression <AttributeReference> high#19 could run on GPU
  @Expression <AttributeReference> collect_set(struct(low, low#17, high, high#19), 0, 0)#24 could run on GPU
  @Expression <Alias> sort_array(collect_set(struct(low, low#17, high, high#19), 0, 0)#24, true) AS sort_array(collect_set(struct(low, high)), true)#25 could run on GPU
    !Expression <SortArray> sort_array(collect_set(struct(low, low#17, high, high#19), 0, 0)#24, true) cannot run on GPU because array expression AttributeReference collect_set(struct(low, low#17, high, high#19), 0, 0)#24 (child StructType(StructField(low,IntegerType,true), StructField(high,IntegerType,true)) is not supported); expression SortArray sort_array(collect_set(struct(low, low#17, high, high#19), 0, 0)#24, true) produces an unsupported type ArrayType(StructType(StructField(low,IntegerType,true), StructField(high,IntegerType,true)),false)
      @Expression <AttributeReference> collect_set(struct(low, low#17, high, high#19), 0, 0)#24 could run on GPU
      @Expression <Literal> true could run on GPU
  *Exec <ShuffleExchangeExec> will run on GPU
    *Partitioning <HashPartitioning> will run on GPU
    *Exec <ObjectHashAggregateExec> will run on GPU. The data type of following expressions will be converted in GPU runtime: buf#28: Converted BinaryType to ArrayType(StructType(StructField(low,IntegerType,true), StructField(high,IntegerType,true)),false)
      *Expression <AggregateExpression> partial_collect_set(struct(low, low#17, high, high#19), 0, 0) will run on GPU
        *Expression <CollectSet> collect_set(struct(low, low#17, high, high#19), 0, 0) will run on GPU
          *Expression <CreateNamedStruct> struct(low, low#17, high, high#19) will run on GPU
      *Exec <FileSourceScanExec> will run on GPU

[NVnavkumar]

See cuDF issue: rapidsai/cudf#11222

[sameerz]

Depends on #7485 and #5430