Is there any reason no to simply use databricks workflow with a continous trigger to consume from a snowflake stream?