__property keyword support in GCC

[Joe Buck]

I wrote:

[ Re: implementation of Borland properties with a helper class ]

> A sufficiently good compiler would optimize this kind of helper class
> object away in any case (though we aren't there yet).

Actually, in this case, there doesn't seem to be a way to get rid of
the extra storage for the pointer, since this helper object would be
a member.

Not that this changes my mind, though, since I think it's better to
show the actual calls to the member functions in the code rather than
try to hide them.