Consistent initialization between all methods

[Robbybp]

All methods we compare need to use "equivalent" initialization. Initializing by strongly connected components is one option, but (a) this is slow and (b) the fullspace flowsheet seems to converge reliably when this initialization is used. Some options:

• Find a different initialization where the implicit function converges (sequential modular?)
• Adjust parameters or ranges so that fullspace has some instances that don't converge even with the initialization

[Robbybp]

Another option is to make the problem more complex, e.g. add a more complicated objective, more bounds, add degrees of freedom (e.g. other flow rates), etc.

[Robbybp]

I'm going to close this, initializing each unit independently seems to meet our needs.