stdlib/String.UnicodeScalarView: fix out-of-bounds read of the

underlying NSString when it ends in a high-surrogate code unit

The tests did not catch this because they were creating CFString, which,
as it turns out, does not perform bounds checking.  Replaced the use of
CFString with a custom NSString subclass.

Swift SVN r19329

stdlib/core/StringUnicodeScalarView.swift

@@ -31,6 +31,9 @@ extension String {
         base = core
       }
       mutating func next() -> UTF16.CodeUnit? {
+        if idx == base.endIndex {
+          return .None
+        }
         return self.base[idx++]
       }
     }