* Add instant iterative development mode. `sparkly-testing --help` for more details.
* Use in-memory db for Hive Metastore in `SparklyTest` (faster tests).
* `spark.sql.shuffle.partitions = 4` for `SparklyTest` (faster tests).
* `spark.sql.warehouse.dir = <random tmp dir>` for `SparklyTest` (no side effects)