branch cut fixup for atanh(bfloat(x))



Personally, I've advocated adding a signed zero (as well as infinities)
to common lisp RATIOs .   It is unfortunate that we appear to
be advocating a retreat from a standard (IEEE 754).  

RJF