Make limits on maximum number of connections in pool.
Release connections after they are used.

This is regarding #79.

It's not fully complete. The limits on the connection pool are currently just magic numbers; I assume you'd want to work that into configuration in whatever standard way you have of doing it.

The tests pass at this point, and the connections do not pile up.

I do, however (as noted in the JIRA ticket) now experience some items not being indexed. I don't know if this is because of problems in my code or if messages are being dropped somehow. I don't tend to see it until I start indexing around 5000-10000 things or so all at once.

I don't know if I'll look further - it's been suggested that we look into using the Camel integration instead. But since I already had this might as well send it along.