Support suggestions

[davidbrochart]

See jupyterlab/jupyterlab#16008.

This PR adds a menu to the Notebook page (it should add it to every shared document in the future), which allows to choose between "Editing" and "Suggesting". The former is the current state (edit document directly), and the later is an editing mode where edits become suggestions, which is the purpose of this PR.
When entering the suggesting mode, the root provider for the shared document is retrieved and a request to fork the document is done with context.model.sharedModel.provider.fork(). This triggers a PUT /api/collaboration/fork_room request with the root room ID of the document in the body. The response gives back a fork ID, which is the room ID for the forked document that is created in the backend. The forked document is a dependency of the root room, and it is kept in sync with the root document: updates to the root document are applied to the forked document, but not the other way around. The shared document is disconnected from the root provider, and connected to a new provider for the forked document. The forked document is advertised on the root Y document as a new state entry fork_{forkId}, so that other users can connect to the forked room. They can do that by calling WebSocketProvider.connect(forkId, sharedModel) on their current provider.
JupyterLab shows a notification when a client enters the suggestion mode, and the user can choose to view the suggestion or remain on the root document. When viewing the suggestion, changes are synchronized with the suggester and it is also possible to collaborate on the suggestion, since it is a regular room. A user (not the suggester) can merge the suggestion or discard it in the Review menu. When merged, a PUT /api/collaboration/merge_room request is done with the fork room ID and the root room ID in the body.
A menu shows which room the document is currently connect to. It can be "Root" for the root room, or the ID of a fork. Users can connect to anyone of them at any time.

[github-actions]

👈 Launch a Binder on branch davidbrochart/jupyter_collaboration/suggestions

[ellisonbg]

@Zsailer @dlqqq for review

[trungleduc]

Thank @davidbrochart for working on this feature. After playing with this PR in JupyterCAD, I like the simplicity of the fork idea and how easy to integrate it into our project. Here are my few feedbacks:

• The hardest point is to keep track of ongoing suggestions. For now, we need to listen to the state-changed event of the YDoc to build the list of suggestions. In our case, we have multiple opened documents with only one view to list the suggestions, it becomes harder to keep track of the suggestion list when switching documents.
• The forked<->root switching mechanism does not work as expected, returning to the root from a forked room brings all changes from the forked room to the root room.
• It's not possible to discard a forked room without merging it.

My wish list for the improvement:

• Suggestion can store metadata like label, author,... It would help to create a rich UI for the suggestion panel.
• Being able to list all suggestions with metadata of a document anytime, delete the suggestion

[davidbrochart]

Thanks for the feedback Trung! You suggestions make a lot of sense, I'll work on them soon.

[davidbrochart]

Closing in favor of #292.