[UMBRELLA] Generalise form 'static' and 'dynamic' risk factor models

[jamesturner246]

For context, in existing HealthGPS speak, static means initialise and dynamic means update.

The concept of the 'static' and the 'dynamic' risk factor models is already a little cloudy.

Some Dynamic models also do initialisation, and vice versa. Both model types have generate_ and update_risk_factors methods. So why not relax the restriction of having static and dynamic models, and just use the generate and update methods of both where appropriate?

This would also allow us to create as many risk factor models as we like, and stop polluting our models with hacky 'extras' where needed.

Also perhaps some way of grouping these modules, akin to the current separation of risk factor, demographic, analysis classes.

Tasks

• Replaced fixed module ordering with ordered list of modules to run #383
• construct tests for new ordered list of module runs #384
• Separate out linear model component of StaticLinear #385
• Construct tests for new linear module in #303 #386

[dalonsoa]

Is this related to #194?

[jamesturner246]

Related, but opposite ideas. @israel-vieira was rightfully against this change because modularity of model components would decrease with such a change -- i.e. models would become more bloated. This change seeks to reverse that by having separate modules for each unit computation -- i.e. the opposite of what #194 proposes. I will close #194, to avoid confusion.