I would definitely prefer a generic Checked ala the one the checked crate provides to the current approach of checked_*, much like the Wrapping type provides a more ergonomic alternative to the wrapping_* methods.
That makes it possible to use core::ops::{Add, Sub, Mul, Div} which also enables polymorphism with respect to the arguments.
Since that doesn't exist though, and NonZeroU* has some of the checked_* methods, filling out the rest makes sense for consistency's sake.
I'm not sure how the type signature would differ other than the rhs operand, which is the u* type mapping to the NonZeroU*.
This gets into one of those reasons I want the full solution to the problem instead. Because arguably NonZeroUN::checked_add could take both uNandNonZeroUN as the RHS. But is it worth doing that? Probably not, because it means another trait (even if hidden). And it'd mean that uN::checked_add should arguably also take NonZeroUN...
If we just had an Integer<MIN, MAX> type, we could avoid all of these questions and problems.