Add spark support

[schuderer]

Closes #8

Add a Spark DataSource (experimental). See examples/spark_datasource.py (see docstrings for documentation and configuration example)

It will create the connection and spark connection object/context for you based on the config.

Use the method get_spark_dataframe() to get a Spark DataFrame, andget_dataframe() to get a Pandas DataFrame.

NOTE: Only use the Pandas variant if you want the data downloaded locally for non-Spark processing. Stick to the spark-specific method to be able to use Spark in further processing and/or ML.

Once you get your hands on a Spark DataFrame, you usually don't need to deal with the spark connection/context explicitly. But if you need it, you can access it through this DataSource's spark property. If you don't need to query data, but need the connection object, you can create a dummy data source (and never call its get_(spark_)dataframe() methods).

Caveats: This being experimental, there is currently no Spark data sink. Let me know if you need it, and I'll take a look at it.