We now have enough support for multiple views for the same action. It's

time to separate out the registration of the models and the views so

there's no longer a one-to-one association.
The new BaseAction.render_model_js() is responsible for writing the

JavaScript to render an action model. Every registered action in Python

will be registered in JavaScript.
The existing BaseAction.render_js() (old name currently kept for

compatibility reasons) will render the view. This is now only called

when the attachment points used on the page, rather than all attachment

points. That's handled in the {% actions_html %} template tag.
The template tag renders the HTML for the attachment point where the

template tag is placed, and then sets up a page injection containing the

JavaScript views. Then, when rendering the JavaScript models, the

injected JavaScript views content will be included.
Some of the design of this change is here to facilitate the upcoming

improvements to attachment points.

Unit tests pass.
Tested that all the actions on the page displayed and worked correctly.