bug#12081: 24.1; buffer-predicate often not called

[martin rudalics]

> Actually, it appears I'm wrong about that.  The existing code doesn't
> make such a guarantee.  Still, it seems to me preferable to have
> replace-buffer-in-windows and replace_buffer_in_windows_safely agree
> on what to do as a last resort; that creates a more consistent user
> experience.

When `replace-buffer-in-windows' fails to replace the buffer,
replace_buffer_in_windows_safely will kick in and automatically do what
you mean.  The user should not notice that.

martin