Create sti.Sim() class

[robynstuart]

Users should be able to circumvent a lot of the sim creation process with a call like

sim = sti.Sim(
    data='./data/', 
    diseases=['ng', 'ct', 'tv', 'bv'], 
    connectors=True
)

[RHull-IDM]

There is a lot to think about here in terms of defaults:

• What should the default network(s) be? Do they vary based on which diseases are selected?
• Demographics?
• Interventions and Analyzers are optional, but are there any that will always be selected with specific diseases?
• Is it reasonable to expect connectors to specify which modules are required by using requires= in the constructor?
• Sim params like pop size, timestep, etc.
• How do we format pars for each of these modules, or do we just require the defaults to use this style of initialization?

[robynstuart]

• demographics: this will be easier once we've resolved Add location parameter starsim#683, but for the moment I think we could do a version like this. We could allow demographics to be (a) None, (b) a list of demographic modules that gets passed directly to the ss.Sim constructor, or (c) a string specifying a location. If someone does this last option we can look for the data to construct the ss.Pregnancy and ss.Death modules in the provided data folder. We will need to document the file naming and format conventions.
• interventions and analyzers: I think these will need to be empty by default, although it would be great to think of a better solution for HIV.
• connectors: Could we allow this to either be a list of connectors or a boolean True, in which case we use default connectors? e.g. if someone has HIV and BV in their module, use that connector. We will need to define this somewhere of course. To me it feels easier to define it in a method of the Sim but if you think it'll be easier to use the requires arguments of connectors that seems reasonable.
• networks: by default add StructuredSexual and Maternal
• sim pars: pass directly

This example should work:

pars = dict(
        dt=dt,
        start=start,
        stop=stop,
)

network_pars = dict(structuredsexual=dict(debut=ss.lognorm_ex(20, 5)))
disease_pars = dict(ng=dict(eff_condom=0.6))

sim = sti.Sim(
    pars=pars,
    network_pars=network_pars,
    demographics='zimbabwe',
    data='./data/', 
    diseases=['ng', 'ct', 'tv', 'bv', 'hiv'], 
    connectors=True
)

However, it should also be possible to pass modules directly, or mix and match:

pars = dict(
        dt=dt,
        start=start,
        stop=stop,
)

demographics=[sti.Pregnancy(), 'deaths']  # Replace the default ss.Pregnancy module with the sti one
networks = sti.StructuredSexual(... )
diseases=[sti.Syphilis(...), 'hiv']

sim = sti.Sim(
    pars=pars,
    networks=networks,
    demographics=demographics,
    diseases=diseases, 
    connectors=True,  
)

[RHull-IDM]

For connectors, I proposed the requires approach because it is dynamic and doesn't require us to maintain a list as things shift. We would populate the list of connectors to use by stepping through all of the connectors, seeing which ones the model config satisfies, then instantiating them.

We certainly could maintain a list of connectors and if we ever have more than one mutually exclusive connector for any disease-disease pair then we may have to do it that way.