[Spark load][Bug] Spark load's column terminator is not consistent with Broker load

[xy720]

Describe the bug
For Broker load:

load label test_db.label13 
( 
    DATA INFILE("hdfs://my-hdfs-host:54310/user/doris/table4") 
    into table test_tbl
    COLUMNS TERMINATED BY "|" 
    (k1,k2,name,clicks)
 ) 
WITH BROKER "doris" 
(
    "username" = "test", 
    "password" = "test"
);

For spark load:

load label test_db.label13 
( 
    DATA INFILE ("hdfs://my-hdfs-host:54310/user/doris/table4") 
    into table test 
    COLUMNS TERMINATED BY "|" 
    (k1,k2,name,clicks ) 
) 
WITH RESOURCE "spark0";

They have the same table schema、same upstream data source. The former(broker load) run successfully, but the latter(spark load) report an error "quality not good enough to cancel". And if you change the column terminator of spark load to "\|" or "\\|", then it will run successfully.

Expected behavior
We should keep the column terminator's format of spark load consistent with that of broker load.