Which pattern is most suitable in Databricks for distributing a dataset to enable parallel processing, including multiple transformations or writing operations to various tables or systems?