Issue Description:

The documentation for SignedNumeric says "A type that can represent both positive and negative values."

That's where the protocol started, but that's not actually what it does; a type should conform to SignedNumeric if it is closed under negation (i.e. if additive inverses exist). So, for example, Complex numbers or elements of a finite field don't really have a notion of "positive and negative", but should conform to SignedNumeric. We need to be careful not to get too technical with the documentation of this protocol (more casually "a type should conform if it is Numeric and you can negate things").