It does exist already as a compiler optimisation method, where the compiler infers possible values at different locations in the code. It's called Value Range Propagation or Value Range Analysis [1]. Loops complicate it as a sibling comment mentions, but they don't prevent it from being useful.

There are indeed an infinite number of similar traits. Compilers have to limit themselves to a reasonable subset, because if you take the idea of general numeric traits far enough, you end up with general purpose theorem proving.

[1] https://en.wikipedia.org/wiki/Value_range_analysis