relates to:

• SR-15228 Inconsistent behaviour of Decimal.significand on Darwin and Linux
• SR-15134 Decimal.init(sign:exponent:significand:) gives incorrect values when exponent overflow or underflows

Issue Description:

The significand of a floating-point value should never be negative. Decimal, despite mimicking FloatingPoint APIs, returns negative significands for negative values:

import Foundation

let x = -42 as Double
x.significand.sign // plus
let y = -42 as Decimal
y.significand.sign // minus (!)

This issue was discovering when trying to fix the implementation of Decimal.ulp, where it would have been nice to be able to use this API but for this incorrect sign.