Considering the smallest claimed layer resolution I know of for 3D printing - 10 microns. that’s about 4 ten-thousandths of an inch. You would need a C2 grade precision-ground ballscrew to achieve that. Unless you buy one surplus and retrofit it into your 3D printer, virtually no home 3D printer has such a ballscrew - it would cost more than the entire printer. The typical rolled ACME screw and anti-backlash Delrin nut used for the Z axis has a lead error 10 times worse than a C2 ballscrew. So you’d need some form of closed-loop feedback like a linear scale, also non-existent on most every single home 3D printer (and most commercial printers.) But even then you’re at the mercy of the stepper motor, which has a positional accuracy of 5% of 1.8deg. on either direction. And micro-stepping is not exactly the most reliable way to gain accuracy. Resolution and accuracy are two different things. Next, the roundness of the part is totally dependent on the resolution of the system - and in most all 3D printers, they use belts which lowers the actual mechanical resolution of the system. Finally, the 3D printer is not working off any geometry - it’s creating point-to-point moves based on a point cloud, so basically a 3D printed curve consists of many straight lines, which end up as facets in a finished part.
That said, the parts look really good for home 3D printing, but I question the accuracy of the optics especially if post-processing involves manual sanding which would invariably change the shape of the optic. There are amateur telescope enthusiasts that grind far more accurate optics entirely by hand.