i386/125011: precision of constants for long double

[Satoshi Kimura]

Hi remko,

> Bruce had replied to this, but no responses came in, so I'd figure the
> problem had been resolved.
 I had read his reply, but his mail and any other documents
could show why this special specification was essential only
on FreeBSD/i386. So I cannot reply his comment nor make any
proposal about this problem. Sorry.

 I agree to close this PR. But I think it is better to show
this specification with its reason to man page of ieee or so.

Sincerely yours.
                    Satoshi Kimura  (satosi.kimura at nifty.com)