Refactor the synth::datasource module in order to ease the addition of new data sources

[juniorbassani]

Currently, working on the synth::datasource module is not straightforward as it contains a lot of duplicated code. The problem has become more evident with the new proposition to add support to SQLite into Synth #185. Observe how the author of the PR had to create even more duplication in order to fulfill their purpose. I think we should consider refactoring said module. This would streamline the addition of new data sources and reduce the likelihood of bugs caused by copy/paste.

On the top of my mind, I would propose the following steps:

1. Achieve consensus on the scope of the traits in the module (DataSource and RelationalDataSource). For example, is it possible to require that implementors of RelationalDataSource simply provide the SQL query to do some task, e.g., retrieve the primary keys, instead of also having to interact with the database gateway?
2. Remove the duplicated code. This should be a direct consequence of item 1.
3. Evaluate the consequences of the refactoring when adding new data sources: if the refactoring is well thought, it is expected that adding new data sources should become a simpler task.

Please, share your thoughts and ideas!