Refactor CTA code so that it can be build without Oracle dependencies

At the moment it is not possible to build CTA without Oracle instantclient dependencies, even if we only intend to use PostgreSQL. This can be a problem because CERN oracle-instantclient RPM cannot be distributed outside CERN (external sites can use Oracle public RPMs, but the version we are using is not guaranteed to be available for them).

As a short term solution for the public release we decided to create a metapackage that quietly manages these dependencies inside and outside of CERN (#1246).

However, as a long term solution, we want to be able to build CTA with or without Oracle dependencies, based on the Catalogue type.