[BUG] Failed to convert string with invisible characters to float

[thirtiseven]

Describe the bug
If a number string starts/ends with invisible characters, Spark can convert to double/float normally, but plugin will return null.

Steps/Code to reproduce bug
“value” column is of StringType, the length is its value is 5, and contains invisible character, the real content is as below:

1234^@   ---\u0000
1000^A   ---\u0001

native Spark:

scala> val df = spark.read.parquet("nullcharacter_test_parquet")
df: org.apache.spark.sql.DataFrame = [value: string]

scala> df.selectExpr("value", "length(value)", "cast(value as float)").show(false)
+-----+-------------+------+
|value|length(value)|value |
+-----+-------------+------+
|1234|5            |1234.0|
|1000|5            |1000.0|
+-----+-------------+------+

spark-rapids:

scala> val df = spark.read.parquet("nullcharacter_test_parquet")
df: org.apache.spark.sql.DataFrame = [value: string]

scala> df.selectExpr("value", "length(value)", "cast(value as float)").show(false)


+-----+-------------+-----+
|value|length(value)|value|
+-----+-------------+-----+
|1234|5            |null |
|1000|5            |null |
+-----+-------------+-----+