fix(spark): CREATE TABLE ... PARTITIONED BY fixes

[barakalon]

Spark 3 supports two CREATE TABLE formats: https://spark.apache.org/docs/latest/sql-ref-syntax-ddl-create-table-hiveformat.html https://spark.apache.org/docs/latest/sql-ref-syntax-ddl-create-table-datasource.html

Some of the spark dialect assumes the former, some assumes the latter, and you can easily end up with an invalid statement.

All this DDL transpilation feels very messy to me... I hated this change.