Rework DragonGen.py

[LorenzoPane]

Flow

The generation process now more closely mirrors the build.ninja output itself. The process looks like this:

1. Variables are generated in

   def generate_vars(var_d: dict, config: dict, target: str) -> ProjectVars:

   The returned object is pretty much exactly what will be exported as variables in the build.ninja file.

2. The variables are plugged into an unevaluated outline: a list of objects each representing a Ninja statement

   def generate_ninja_outline(variables: ProjectVars) -> list:

   and extended with the rule and build portions of the outline from

   def build_statements_and_rules(variables: ProjectVars) -> (list, list):

3. The outline is evaluated in with the variables and written to a buildgen generator (which can be easily swapped out) in

   def generate_ninja_file(outline: list, variables: ProjectVars, stream: TextIO):

   Added classes

   To minimize string manipulation most data is stored as ArgList, BoolFlag, or str objects which are properly converted to strings during evaluation.

Conclusion

I hope this will help with maintainability. Since pretty much everything is iterative, most support additions should just involve dictionary modification.