Build ASTree driven by a data structure instead

[gunnarx]

• Consider if there's a way to build the AST from a configuration data structure (reflecting the schema). As of now it is done very 'directly' in concrete code that references specific functions for each data type, etc. An alternative programming style is to write a "generic" algorithm that is instead driven by an input data structure.

This is an idea, and might look more "elegant" than the very direct approach in current parser.py. On the other hand, the direct approach ought to be easily readable and understandable. Simplicity might not be bad from that perspective. This is mostly something to think about, and not critical because the current code "just works".

[gunnarx]

Proposed in #26