Dynamic inclusion of node types

[dalonsoa]

In the current implementation, Model.nodes_types is populated when you create a Model object based of the models that are known to WSIMOD at that time, and regardless of whether those nodes are needed or not for the simulation. As a result:

• Adding custom nodes, as done in Enable customisation #54 becomes more complicated as this needs to be modified manually.
• Some parts of the code are run that are might not be needed, as there are no nodes of that type.

This PR takes a few steps to make this more flexible.

[barneydobson]

This seems sensible. As part of this would it be sensible to stop using the class.name as part of the model behaviour and instead make that a default node attribute? I had made this poor decision super early on and since realising that it's not a good idea it has now become all over the place. Maybe it is a separate issue because it seems to me a reasonably large task and the documentation here would need to be updated.

[barneydobson]

OK and further I have tried this branch on one of our commonly used models and am getting an error that I assume is related to the new registry:

I haven't tested the CLI yet so maybe I am using it wrong (although it works fine on the main branch, you might see some error warnings but these are negligible) - hopefully you can reproduce with below.
model.zip

[dalonsoa]

Ok, I've updated the code - it was just a typo in one place - and now the model works beginning to end without errors. Could you check if the actual numbers produced are still correct?

About your comment, yes let's make a separate issue.

[barneydobson]

I'll test it out on one of our terribly complicated models to see if results are the same

[barneydobson]

I have tested for the Lea model and results are identical. I think that's sufficient for me!