Proposal for Deprecating Signaling NaNs

[David G Hough 754R work <754r@xxxxxxxxxxx>]

As an outcome of today's meeting, I wrote a proposal for what we didn't quite
get to discussing before time ran out.

http://754r.ucbtest.org/proposals/snan/snan.pdf

This is a refinement of previous proposals in that the normative text of
the draft calls Signaling NaNs optional and the Appendix N says they should
not be universally implemented in the 754 way raising an invalid exception
but should be provided where appropriate in an implementation-dependent way
as part of a complete implementation-specific and application-specific
solution to some problem of end-user interest.