Initialization order of Widget vs Bindings

[jdaignault]

We encountered an interesting scenario where we had the following setup:

• Within a widget blueprint graph, an Event Construct was fetching a Dynamic Material from an Image widget and cache it to a blueprint variable
• Within a widget binding, it would call a blueprint function that tried to access the cached dynamic material and update a material parameter

As it turns out, when the binding executed for the first time, the cached variable turned out to be invalid. After poking around in UUserWidget::NativeConstruct, it was observed that Extension->Construct takes place before the widget's Construct blueprint event.

This scenario was easily fixed by moving the event graph logic to Event PreConstruct, but it raised a question about whether this initialization order seemed correct. Should bindings only try to initialize themselves only after widgets have 'fully initialized' themselves?