[stdlib] Teach abs(_:) to use magnitude when it can.

Within the (single) implementation of abs(_:), dynamically check whether
the numeric type and its `Magnitude` are of the same type and, if so,
return the result of `magnitude`. This ensures that we do the right thing
with respect to (e.g.) floating point values like -0.0, without resorting
to overloading of abs(_:).

Author: DougGregor

stdlib/public/core/Integers.swift

@@ -333,6 +333,10 @@ extension SignedNumeric {
 /// - Returns: The absolute value of `x`.
 @inlinable
 public func abs<T : SignedNumeric & Comparable>(_ x: T) -> T {
+  if T.self == T.Magnitude.self {
+    return unsafeBitCast(x.magnitude, to: T.self)
+  }
+
   return x < (0 as T) ? -x : x
 }
 

test/stdlib/FloatingPoint.swift.gyb

@@ -413,6 +413,11 @@ extension FloatingPoint {
   public var isQuietNaN: Bool { return isNaN && !isSignalingNaN }
 }
 
+func genericAbs<T: SignedNumeric & Comparable>(_ x: T) -> T {
+  return abs(x)
+}
+
+
 %for Self in ['Float', 'Double', 'Float80']:
 % if Self == 'Float80':
 #if !os(Windows) && (arch(i386) || arch(x86_64))
@@ -533,6 +538,12 @@ FloatingPoint.test("${Self}.significandWidth") {
   expectEqual(0, ${Self}(2).significandWidth)
   expectEqual(1, ${Self}(3).significandWidth)
 }
+
+FloatingPoint.test("${Self}.absNegativeZero") {
+  expectEqual(abs(${Self}(-0.0)), ${Self}(0.0))
+  expectEqual(genericAbs(${Self}(-0.0)), ${Self}(0.0))
+}
+
 % if Self == 'Float80':
 #endif
 % end