Refactor: clean up codebase, simplify where possible

[z3z1ma]

This will be a WIP for a bit. dbt has evolved enough since this tool was created that we can probably simplify some things. Furthermore, certain experimental things can be removed. I still like the workbench. Suffice to say there are interesting streamlit applications we have not explored like an interactive documentation input flow for progenitors, etc.
But otherwise a lot we can kill. We can also remove core interface folding it into osmosis.py and leverage whatever dbt core has available now in its pythonic interface -- and be more strict on what dbt version(s) are supported with a focus on newer versions.
Also includes cleanups with py 3.9 being min version.

[syou6162]

Also includes cleanups with py 3.9 being min version.

I'm sorry if there is a conflict between the changes made to #190 and what you are trying to do with this pull request 🙏🙏🙏 .

[z3z1ma]

@syou6162 there is some more to do, but at least you can see the significant changes here which will simplify maintainability and even extensibility for LLM integrations, user defined plugins / column fuzzing logic, etc...

My approach in such a refactor as this is to collapse code into one file as much as possible such that we can centrally reason on the abstractions and code paths. From there, I can re-break out the functions into different namespaces/modules (if really needed) once the codebase refactor is done. The tests will be updated as appropriate soon too and hopefully things reveal themselves as easier to test as well which would evidence we move in the right direction.