Evaluator rework - propagate flags through evaluation

Reopened to have place for documentation, currently not ready for revision!

This is (another) larger rework of the evaluator motivated by #18 (closed). This issue will be fixed, once this is merged and as an intended and appreciated side effect, all the groundwork for a generic data processing function a la flagGeneric is done.

The other major change is organizational: The AST checking and rewriting are now separated into different tree walks and evaluator.py was split up into an new core-submodule.

Wanna have a look, @palmb ?

BTW: No squash please.