Coordinate closing continuous executions across windows and data sources

From discussion https://community.jboss.org/thread/201379?tstart=0

In a continuous execution involving more than one data source, it is possible that one or more data sources will reach end of data with no possibility of more results. The translator API should allow for signalling this to the query engine. Further for the query engine ought to extend non-continuous end of result set processing semantics to the continuous execution case as much as possible.

Below is a stab at what the continuous execution end of result set semantics might be.

There are two cases: correlated results across data sources and non-correlated results.

In the correlated case (such as an inner join), when one data source indicates end of results, the execution ends for all correlated data sources.

In the uncorrelated case (such as a union), any data source that signals end of results is not dropped from the plan on the next window and its resources are cleaned up.

In both cases when all data source executions end, the client should be notified through a callback mechanism.