Fix powf().

Summary:
From Steve Kargl:

Paul Zimmermann has identified a bug in Openlibm's powf(),
which is identical to FreeBSD's libm.  Both derived from
fdlibm. JuliaMath/openlibm#212.

Consider

% cat h.c
int
main(void)
{
  float x, y, z;
  x =  0x1.ffffecp-1F;
  y = -0x1.000002p+27F;
  z =  0x1.557a86p115F;
  printf("%e %e %e <-- should be %e\n", x, y, powf(x,y), z);
  return 0;
}

% cc -o h -fno-builtin h.c -lm && ./h
9.999994e-01 -1.342177e+08 inf <-- should be 5.540807e+34

Reviewers: manu

Subscribers: imp, andrew, emaste

Differential Revision: https://reviews.freebsd.org/D31865

(cherry picked from commit 292815e)

Author: Mark Murray

lib/msun/src/e_powf.c

@@ -136,7 +136,7 @@ __ieee754_powf(float x, float y)
     /* |y| is huge */
 	if(iy>0x4d000000) { /* if |y| > 2**27 */
 	/* over/underflow if x is not close to one */
-	    if(ix<0x3f7ffff7) return (hy<0)? sn*huge*huge:sn*tiny*tiny;
+	    if(ix<0x3f7ffff6) return (hy<0)? sn*huge*huge:sn*tiny*tiny;
 	    if(ix>0x3f800007) return (hy>0)? sn*huge*huge:sn*tiny*tiny;
 	/* now |1-x| is tiny <= 2**-20, suffice to compute
 	   log(x) by x-x^2/2+x^3/3-x^4/4 */