I'd argue in favor of those. A competent text editor (let's say no older than 40 years old) can highlight and/or render that. When using base markdown, those things add structure to the document. I'd much rather see a markdown file with basics like headers and code sections (which are usually highlighted correctly) than a raw .txt file.

however I think your argument goes doubly hard when dealing with these more advanced html integrated readmes. The md file of the example readme is not well readable when opened in a text file to me- and there's actually very little content there! I prefer a more minimal approach where opening the .md file gives me the information I need, and opening it in a rendered environment gets me some niceties (perhaps you've got badges with coverage or versions nrs idk). But it should work in both.