Not sure if I agree with this. Ideally, this would be the case, but the tooling and dominant platforms just aren't as mature as they would need to be to consider it a required core skill. Doing Accessibility work feels like doing CSS work back in the late 90s. The dominant target platform, JAWS, is just a horribly coded product. It kinda, sorta, handles ARIA attributes and it kinda, sorta, does the same thing each time a page is loaded, but then also doesn't.
And it's a very expensive tool. The "pro" version, which is the only version approved for dev work is $1100 (outside of the extended 90 trial for $179, I guess). Now consider that not only do your devs need copies, but so do your testers. And if they want to stay up to date, you're really looking more like $1100 + $550 a year. And, yes, there's other options, but since a lot of accessibility work is really more for lawsuit avoidance than anything else, not going with the industry standard isn't a wise choice.
And, after all of that, unless you work someplace with enough funds to properly test out your product with, you know, actual users with visual impairment, what qualifies as good accessibility is difficult to suss out. You can run automated tools, but they're just code linting tools, more or less, and really speak nothing to the user experience. How do you really test out the UI you've developed? If you close your eyes or cover the screen you still know exactly what it looks like.