[BUG] Handle regexp_replace inconsistency from https://issues.apache.org/jira/browse/SPARK-39107

[NVnavkumar]

Describe the bug
When passing an empty string ('') and a regular expression containing only ? or * repetitions, the output is not consistent between Spark and cuDF. Note this is actually a pure inconsistency with Spark's regexp_replace, it actually does not apply to what Java does as a standard.

Steps/Code to reproduce bug
PySpark Example Code:

from pyspark.sql import SparkSession

spark = SparkSession.builder \
      .master("local[*]") \
      .appName("regex_empty_match") \
      .config("spark.rapids.sql.regexp.enabled", "true") \
      .config("spark.rapids.sql.explain","ALL") \
      .config("spark.rapids.sql.castStringToTimestamp.enabled","true") \
      .config("spark.rapids.sql.exec.CollectLimitExec", "true") \
      .config("spark.rapids.sql.castFloatToString.enabled", "true") \
      .config("spark.rapids.sql.hasExtendedYearValues", "false") \
      .getOrCreate()  

df = spark.sparkContext.parallelize([[""],[""],["AAA"]]).toDF(["a"])
df.show()

df.selectExpr("regexp_replace(a,'A*','_REPLACED_')").show()

Spark (CPU) Output:

+------------------------------------+
|regexp_replace(a, A*, _REPLACED_, 1)|
+------------------------------------+
|                                                               |
|                                                               |
|                _REPLACED__REPLACED_  |
+------------------------------------+

Plugin (GPU) Output:

+-------------------------------------+
|regexp_replace(a, A*, _REPLACED_, 1)|
+-------------------------------------+
|                          _REPLACED_                |
|                          _REPLACED_                |
|                _REPLACED__REPLACED_   |
+------------------------------------+

Expected behavior

Empty string input should be short circuited in the plugin as this is what is expected in certain versions of Spark.

Additional context

This bug was originally reported to Spark in https://issues.apache.org/jira/browse/SPARK-39107, and the issue was fixed in apache/spark#36457 for newer patch versions of Spark in 3.1,3.2,3.3 and master, so shims will need to be created to handle the original faulty behavior.

[anthony-chang]

Depends on #5450 being merged first