Add UTF-16 string literal constructors to String classes

third_party/WebKit/Source/wtf/text/WTFString.h

Index: third_party/WebKit/Source/wtf/text/WTFString.h
diff --git a/third_party/WebKit/Source/wtf/text/WTFString.h b/third_party/WebKit/Source/wtf/text/WTFString.h
index 354001358ba11f6bde98c02a52e155ac3e3f3d57..3c516adf647f803ba4eb9509c21688d456bb787d 100644
--- a/third_party/WebKit/Source/wtf/text/WTFString.h
+++ b/third_party/WebKit/Source/wtf/text/WTFString.h
@@ -84,6 +84,8 @@ class WTF_EXPORT String {
 
   // Construct a string with UTF-16 data, from a null-terminated source.
   String(const UChar*);
+  String(const char16_t* chars)
+      : String(reinterpret_cast<const UChar*>(chars)) {}

[jungshik at Google, 2017/03/09 20:33:33]

When the C++ spec is strictly interpreted, this casting would lead to an
undefined behavior with some optimization mode. 

https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html. See
-fstrict-aliasing. 

Anyway, this adds an interesting wrinkle to ICU update to 59(to-be), which
changes UChar to char16_t.  

For non-ICU codes, UChar can be configured to what used to be and it works fine
with just one change in ToT blink. However, it leads to a goma failure. See
https://codereview.chromium.org/2738453005/ (v8) and
https://codereview.chromium.org/2740673002/ (chromium/blink ). 

So, I tried to make UChar always char16_t within and outside ICU and multiple
definition of ctor errors and found this change. 

 
   // Construct a string with latin1 data.
   String(const LChar* characters, unsigned length);