GeoServer works with memory limits, which are a more accurate and predictable way to control memory usage.
The image size per se is not very meaningful, depending on bit depth it can be very small, like with packed black and white images storing 8 pixels in one byte, or large, with up to 4 bytes per image (assuming 8 bit outputs). Also, image size it's too dependent on aspect ratio (a certain output can fit the total amount of pixels but fail the check because its aspect ratio is not the same as the limits one).
The downside of memory limits is that they cannot be advertised in the capabilities document. The beauty of open source is that you can push for GeoServer to implement that feature, follow this guide if you are interested: