This carries an "aint nuffin clever in 'ere" warning: So far All I have done
  is get yasvec compiling again, remove a few bugs and extend reallocate
  for some cases where msize and expand are availible.
  
  enable_if removed, its function replaced with type_selector in the case
  of namespace level functions and the use of dummy Int2Type arguments
  in member functions. I don't like dummy arguments but they seemed better than
  the other options (namely external, explicitly templated functions 'hidden' in
  an implementation namespace).  

  I should have kept a log of all of the bugs I pulled out and other modifications,
  Indeed, I almost did but somehow it didn't quite happen - sorry. 
  
  uninitialised_copy I have defined for InputItr = const T* and InputItr = T* 
  seperately, this is less than desireable but previously with only the 
  const-data version defined any non const pointers were being eaten up by
  the default templated function. This is fairly minor but I mentions 
  because you may disagree with me.
  
To do:
      Cut vector back vector to a buffer then decide whether it is sensible to 
      re-implement vector using this buffer. 
      
      Implement valarray and related classes using yasli_nstd::buffer and other
      yasli components. The structure of these classes will follow that found in 
      Mingw32 - expressions are to be evaluated during assignment. 
      
      Add yasli_nstd::reference_valarray, a class which behaves like 
      valarray and can be used in conjuction with it but which consists of an
      array of pointers.