[BUG] Approx percentile query fails with UnsupportedOperationException

[andygrove]

Describe the bug
Executing an approx_percentile query against a single row of input and with no group by clause causes an UnsupportedOperationException.

Steps/Code to reproduce bug

from pyspark.sql.types import *
schema = StructType([StructField("byteF", ByteType()), StructField("shortF", ShortType())])
data =[(1, 2)]
df = spark.createDataFrame(data, schema)
df.createOrReplaceTempView("t")
spark.conf.set("spark.rapids.sql.expression.ApproximatePercentile", "true")
spark.sql("SELECT approx_percentile(byteF, 0.75) FROM t").show()

Fails with:

java.lang.UnsupportedOperationException: approx_percentile does not support reduction
	at com.nvidia.spark.rapids.GpuApproximatePercentile.initialValues$lzycompute(GpuApproximatePercentile.scala:69)

Expected behavior
Should not fail.

Environment details (please complete the following information)
Spark 3.1.1 local mode

Additional context
None

[GaryShen2008]

Correct the project to 22.04.