Move as much logic out of NGen.cpp as possible

[christophertubbs]

NGen.cpp defines the main function, along with many other tasks, such as:

1. The simulation loop
2. Layer construction
3. Routing
4. Timing and reporting
5. Output manager construction
6. Feature Construction
7. Formulation creation and reading
8. Partition handling
9. Hydrofabric loading
10. File validation
11. python interpreter creation
12. MPI setup and teardown

NGen.cpp and its main function should, ideally just be providing an entry point from the CLI. This setup limits the growth and flexibility of NGen as it requires the use of main to launch and run the operations that allow NGen to work in the correct order with the correct wrapping logic.

[PhilMiller]

I'm about to post changes that make up a good chunk of this effort. It's based on changes I and others (mostly Ian Todd) did under NGWPC to support state saving, but aimed at generalizing the de-coupling of many of these things.

[christophertubbs]

Excellent. Did you by chance run int any issues regarding state management? It's currently a vague concern, i.e. "Is this going to be a problem and what will we need to do?"

[PhilMiller]

Surprisingly few, actually. There's still plenty of work to do in terms of demonstrating it in real usage, though (hot-start, calibration, retrospective, hindcast, AnA, etc)