This PR

• resolves Need for signbit in the array API spec #670 by adding a specification for signbit for determining whether a sign bit is set for each floating-point number in a provided input array.
• follows Python conventions in which a boolean is returned. In C99, this is a macro which returns an integral value, where a positive result can be any non-zero integral value and a negative result must be zero.
• limits the input data type to real-valued floating-point data types. The current text uses "should", indicating that array libraries may extend support to additional data types (e.g., signed integer data types) for, e.g., backward compatibility reasons. This PR does not use the more liberal real-valued data type, as the intent of this function is for distinguishing +0 and -0 and the determining the sign of NaN, values which are not possible outside of floating-point data types. For complex data types, users need to provide the real and imaginary components separately.
• as this API is commonly implemented across array libraries without deviation, there are no open questions requiring resolution before specification inclusion.