There’s two approaches that can be used here, general and particular.
For scaling the images generally, the content management system can include a post-processing rule regarding how the dimensions HTML is generated. This approach is adquate, maybe even preferred, for a number of systems where you’d want the system (instead of the author) to take care of how the images are scaled automatically (e.g. making all of the images in a news article or forum topic the same width).
For scaling particular images to particular dimensions, I agree some kind of additional syntax would be required. Since people are wanting to add other particular attributes to page elements in Markdown, this is a good fit for the consistent attribute syntax extension, rather than creating a seperate syntax just for image dimensions. Different scaling attributes/techniques could be supported via different attributes too.
I don’t believe support for image dimensions needs to be added to the core spec, since the general approach + HTML for particulars has worked fine up until now and the already large CommonMark core is focused on the features found in the original Markdown spec. So I propose we move this discussion to “Extensions”.