[FEA] Set spark.rapids.sql.allowMultipleJars to NEVER by default

[thirtiseven]

Is your feature request related to a problem? Please describe.
#9654 Added a new config spark.rapids.sql.allowMultipleJars, which is used to detect if there are multiple plugin/cudf/jni jars on the classpath when initializing RapidsDriverPlugin and RapidsExecutorPlugin and throw an exception if there are.

Currently it's default value is set to SAME_REVISION, which means that jars with the same revision are allowed by this detection. Ideally, we should set it to NEVER by default to avoid the following problems:

• It does not detect when the dist jar is in multiple extraClassPath, $SPARK_HOME/jars, and on the user classpath via packages and jars. We have seen bugs in such deployments [BUG] RapidsShuffleManager and external packages for Spark Standalone #5796
• when you have sql-plugin and aggregator on the classpath at the same time the user might be using an unintended flatbuffers dependency, the aggregator relocates flatbuffers to an internal package through the shade plugin to make sure it uses just the one from the aggregator.

Additionally, we should also provide a way to detect submoudle jars in the classpath, like adding a new config mode RECURSIVE to find out the path of submoudle jars.

Describe the solution you'd like
The reason we didn't set the default value to NEVER is that it may require some breaking changes in CI, to solve it, we can:

• Find any multiple jar usage from the log, and try to remove them.
• Set spark.rapids.sql.allowMultipleJars=NEVER in CI jobs and fix failure if any.
• Set default value to NEVER in code.