PostgreSQL writer

Current implemented for streaming sources as a `foreachWriter`, but would be more general if it were actually a sink.

Sources of inspiration:

* https://github.com/phatak-dev/spark2.0-examples/blob/master/src/main/scala/com/madhukaraphatak/examples/sparktwo/datasourcev2/MysqlWithTransaction.scala
* https://www.slideshare.net/datamantra/understanding-transactional-writes-in-datasource-v2-109643509
* http://blog.madhukaraphatak.com/spark-datasource-v2-part-6/

Start by locating the current PostgreSQL implementation for streaming sources and its foreachWriter integration. Read the linked Spark DataSource V2 transaction examples, then define completion as replacing or generalizing the writer so it functions as a sink rather than only through foreachWriter.

Written by the indexing model from the issue text.