tl;dr:
Why is cursor position not antialiased by applications, desktop environments, or operating systems? Why don't user interfaces give the ability to localize pointer position with subpixel precision?
Pixels are physical things on our physical screens, and are ultimately are the "quanta" of display resolutions. However, conceptually, pixels are not tessellating squares. They are instead discrete, infinitesimal points in two-dimensional space, similar to how "samples" in digital audio are a discrete signal which nonetheless describe a smooth, continuous one-dimensional curve. Consequently, there is no reason to not conceptually think about the "signal" that exists in-between the pixels. In fact, pixel density does not define a spatial localization precision at all; pixel density only limits the highest "frequency" of a longitudinal signal that can exist in a given length.
Indeed, we make use of interpixel space all the time with antialiasing filters! Examples include text1 and geometry rendering2, as well as graphics:
So why can't we do this to cursor position, too?
1Yes, different displays may have different subpixel geometries, which is taken advantage of by text renderers, but that doesn't prevent *pixel-level* antialiasing. Regardless of the subpixel layout, cursor movements can be localized to subpixel precision so long as the whole pixels are laid out in the same grid formation.
2Yes, display gamma complicates antialiasing, but this doesn't prevent antialiasing from being used elsewhere.