Added example task 'Interactive Image Generation'

[meta-paul]

Added a new example of a Model-in-the-Loop task that's based on FormComposer.

The task is an iterative prompt-driven image generation where Worker interacts with the server after every round of responses, and each response is form-based (and represented by a fieldset).

FormComposer was enhanced to support:

• dynamic fieldsets (so multiple instances of the same response fieldset can be saved)
• additional context for triggers to modify react state of the Task form
• buttons as an element of a form fieldset (in addition to the standard form fields)
• more modular set of React components (a refactor of existing larger code blocks)