Update to metadata that fixes the char* as ushort* problem

[AArnott]

No description provided.

[tannergooding]

This may not be a good idea. There is some discussion in dotnet/corefxlab#2350 about changing char to be UTF8 and therefore 1-byte, rather than 2 at an opt-in level which means that char* for a LPCWSTR would now be incorrect.

Likewise, while char* itself is just a pointer and therefore blittable. char, char[], and string are not and are only sometimes blittable (and generally require you to explicitly annotate that you want it handled as UnmanagedType.U2 or as CharSet.Unicode).

[AArnott]

I am shocked to learn that char might be redefined as a 1-byte value. I can't imagine how much code that would break.
I'm not sure what you're saying isn't a good idea though. char* is how anyone and everyone can represent a UTF-16 string today. string can be pinned today as a char*. It cannot be pinned as a ushort*. So while I don't have a crystal ball to foresee what the CLR may do in the future, we're generating code that needs to work well with the code people have today, which for the time being is char*.