Consider ctti::unnamed_type_id()

[Manu343726]

There is some people complaining about string storage, and I actually agree that the fixed size array is not the best thing to do. I have to try again some alternatives with VS.

However, I noticed that when working with ctti for type indexing, ignoring the compile-time demangling-like functionallity, you don't care about type names at all. So storing the hash in type_id_t should be enough.

So I'm thinking about adding a ctti::unamed_type_id_t (and its ctti::unnamed_type_id() counterpart) to just do type indexing, without all the storage needed for the string. See Manu343726/siplasplas#5 for an example of such a case.