[BUG] regexp_extract doesn't work correctly with concat

[sperlingxx]

Describe the bug
regexp_extract(concat(....), ) produces incorrect results in GPU runs, when children of concat contains at least one column vector.

Steps/Code to reproduce bug

val df = (1 to 10).toDF("a")
spark.conf.set("spark.rapids.sql.regexp.enabled", "true")
df.coalesce(1).select(regexp_extract(concat(col("a"), lit("a")), "(a)", 1)).collect()

GPU results: Array([], [], [], [], [], [], [], [], [], [])
CPU results: Array([a], [a], [a], [a], [a], [a], [a], [a], [a], [a])

For above query, GPU works correctly only when column(a) outputs an empty string.

[revans2]

This has nothing to do with concat. It appears that regexp_extract on the CPU is doing a find, where as on the GPU it is doing a full match.

val df =Seq("1a", "2a", "3a", "4a", "5a", "6a", "7a", "8a", "9a", "10a").toDF("c")
df.coalesce(1).select(regexp_extract(col("c"), "(a)", 1)).collect()

shows the same results, but changing the regular expression to be ".*(a).*" produces the same result for both the CPU and the GPU.

[sperlingxx]

Close this issue since it is included in #5135