characterset support changes

[Richard Frith-Macdonald]

I have implemented the latest MacOS-X NSCharacterSet API ... which 
extends the class to support the full unicode characterset rather than 
just the base multilingual plane.

This increases the potential size of a characterset by a factor of 
sixteen, and while there is some attempt to keep the memory consumption 
down (only allocating bitmap memory for a map extending up to the plane 
containing the highest values character in the characterset) I'm 
somewhat concerned that the memory usage of this implementation may be 
excessive.

I'd therefore be interested if anyone using the NSCharacterSet code 
could test this out and see if they think the implementation is OK in 
practice (and of course report any bugs).