Improve App-Schema connection usage

As noted in this mailing list discussion, there's room for improving the connection usage of App-Schema data stores.

Areas of improvement include:

1. Connection re-use (Major) building a feature type containing nested features currently entails opening a separate connection for each joined type, which can go as far as exhausting the connection pool when deeply nested hierarchies are modeled in the application schema
2. Connection leaks (Minor) there are a few spots where a feature iterator is may not be properly closed if an unchecked exception is thrown while iterating, leading to connection leaks