Add support for org.apache.spark.sql.catalyst.expressions.ArrayPosition [databricks]

[ustcfy]

Closes #5224

This PR adds support for org.apache.spark.sql.catalyst.expressions.ArrayPosition.

Ran a simple perf test for 5 times

// use big data gen
import org.apache.spark.sql.tests.datagen._
val dataTable = DBGen().addTable("data", "a array<byte>, b byte", 10000000)
dataTable.toDF(spark).write.mode("OVERWRITE").parquet("PERF")

// spark-rapids
val df = spark.read.parquet("PERF")
// `key` is a scalar
spark.time(df.selectExpr("max(array_position(a, 0))").show())
// `key` is a column
spark.time(df.selectExpr("max(array_position(a, b))").show())

Results:

• key is a scalar

  	Run 1	Run 2	Run 3	Run 4	Run 5
  GPU	208 ms	209 ms	263 ms	265 ms	225 ms
  CPU	898 ms	985 ms	996 ms	953 ms	868 ms

• key is a column

  	Run 1	Run 2	Run 3	Run 4	Run 5
  GPU	212 ms	237 ms	220 ms	211 ms	221 ms
  CPU	956 ms	994 ms	892 ms	914 ms	932 ms

The results showed that the GpuArrayPosition (TITAN RTX) demonstrated approximately 4 times speedup compared to the CpuArrayPosition (8 CPU cores).

[gerashegalov]

Anonymous classes are unnecessarily difficult to debug when the conversion code throws #10838

Look at https://github.com/NVIDIA/spark-rapids/pull/10839/files for an example

[firestarman]

Looks good to me

[ustcfy]

build