Add ability to specify dataset in connection string #24

We have a use case where we want to treat separate datasets as similar to separate databases or schemas in a traditional database. This would allow us to have separate bindings to different datasets.

An example of the kind of behavior we'd like:

engine = create_engine('bigquery://some-project-name/some_dataset')

This is a pretty important use case for us, since it would make a multi-tenant approach much easier.