This seems like an API failure that common jobs for extensions can't be done without scary permissions.
edit: since content blockers in particular are so popular, it might even make sense for them to create a security-conscious content blocker specific API like Apple did
Google’s not rushing to implement this for obvious reasons. They will most likely continue to push AMP and at some point decide to block “malicious ads” or whatever, so that in one move they can both cut out the competition (by being the de facto arbiter of acceptable ad technology) and make sure they don’t negatively affect their business.
It’s still a net positive for the user, at least in the short term.
Adding, or modifying any DOM node, to the effect that it includes third-party resources could be denied.
If you also deny adding inline <script> blocks and enforce any injected javascript to run in strict CSP mode it could potentially work.
The problem is that being able to modify the page is a method of exfiltrating information - e.g. you can add an img tag triggering a GET to an arbitrary URL. I suspect it would be surprisingly tricky to try restricting page modification in such a way that information can't be sent anywhere the page wouldn't have sent it originally.
But +1 on deferring permissions prompts until the moment the app actually needs the privilege, though. It's worked pretty well on iOS, and it's finally coming to Android - maybe it'll come to Chrome or Firefox next.