Retrieve dbname in schema queries, fill the catalogs.

At the moment the dbname is just duplicated information on each row. In the
future we will be able to implement a new --all-databases switch to support
looping over all databases of a single cluster and share COPY resources
between the databases in a smart/global way.