Shapes in Action Discovery

[calummackervoy]

Actions should publish their inputs with Shapes, returned during Action Discovery. This way the front-end can know whether it has the necessary information for an action, and can make the request without implicit knowledge of what the action is

I suppose that for now the description of an Action's output can just be a text description. Later it could simulate the effects of the action and return the graph changes for the front-end to display