Visual editor: Allow to edit in-place elements containing more than one text node

Right now editing in place (using contenteditable, introduced in #121) is only allowed if the element whose content is being edited contains at most one text node (and any number of non-text nodes, like HTML element), but as soon as there are more than one text node we can't know which of the nodes has been edited (or if both were!)

Right now those can only be edited on the sidebar.

For a nicer user experience this should be possible