Spark unbounded source advances watermarks prematurely

`SparkUnboundedSource` will advance the watermark to the MAX of the watermark of any partition. You can see it at [https://github.com/apache/beam/blob/fab12c772d461fc8db4b3c361d38fe2781926fff/runners/spark/src/main/java/org/apache/beam/runners/spark/io/SparkUnboundedSource.java#L204](https://github.com/apache/beam/blob/fab12c772d461fc8db4b3c361d38fe2781926fff/runners/spark/src/main/java/org/apache/beam/runners/spark/io/SparkUnboundedSource.java#L204) .

This should be the MIN - this is a combining of watermarks - not advancing. This currently means the watermark moves too quickly and the slowest partition of an unbounded source has elements that are routinely marked late.

Imported from Jira [BEAM-7423](https://issues.apache.org/jira/browse/BEAM-7423). Original Jira may contain additional context.
Reported by: mikekap.