math: fix ld80 acoshl(x) for x < 0

acosh(x) is nan for x < 1, but x < 0 cases were not handled specially
and acoshl gave wrong result for some -0x1p32 < x < -2 values, e.g.:

acoshl(-0x1p20) returned -inf,
acoshl(-0x1.4p20) returned -0x1.db365758403aa9acp+0L,

fixed by checking the sign bit and handling it specially.

reported by Paul Zimmermann.

src/math/acoshl.c

@@ -10,14 +10,18 @@ long double acoshl(long double x)
 long double acoshl(long double x)
 {
 	union ldshape u = {x};
-	int e = u.i.se & 0x7fff;
+	int e = u.i.se;
 
 	if (e < 0x3fff + 1)
-		/* |x| < 2, invalid if x < 1 or nan */
+		/* 0 <= x < 2, invalid if x < 1 */
 		return log1pl(x-1 + sqrtl((x-1)*(x-1)+2*(x-1)));
 	if (e < 0x3fff + 32)
-		/* |x| < 0x1p32 */
+		/* 2 <= x < 0x1p32 */
 		return logl(2*x - 1/(x+sqrtl(x*x-1)));
+	if (e & 0x8000)
+		/* x < 0 or x = -0, invalid */
+		return (x - x) / (x - x);
+	/* 0x1p32 <= x or nan */
 	return logl(x) + 0.693147180559945309417232121458176568L;
 }
 #elif LDBL_MANT_DIG == 113 && LDBL_MAX_EXP == 16384