gh-121889: cmath.acosh(0+nanj) returns nan+pi/2j (#121892)

As per C11 DR#471 (adjusted resolution accepted for C17), cacosh (0 +
iNaN) should return NaN ± i pi/2, not NaN + iNaN.  This patch
fixes cmath's code to do same.

Author: skirpichev

Lib/test/mathdata/cmath_testcases.txt

@@ -371,9 +371,9 @@ acosh1002 acosh 0.0 inf -> inf 1.5707963267948966
 acosh1003 acosh 2.3 inf -> inf 1.5707963267948966
 acosh1004 acosh -0.0 inf -> inf 1.5707963267948966
 acosh1005 acosh -2.3 inf -> inf 1.5707963267948966
-acosh1006 acosh 0.0 nan -> nan nan
+acosh1006 acosh 0.0 nan -> nan 1.5707963267948966         ignore-imag-sign
 acosh1007 acosh 2.3 nan -> nan nan
-acosh1008 acosh -0.0 nan -> nan nan
+acosh1008 acosh -0.0 nan -> nan 1.5707963267948966        ignore-imag-sign
 acosh1009 acosh -2.3 nan -> nan nan
 acosh1010 acosh -inf 0.0 -> inf 3.1415926535897931
 acosh1011 acosh -inf 2.3 -> inf 3.1415926535897931

Misc/NEWS.d/next/Library/2024-07-17-09-23-03.gh-issue-121889.6se9jS.rst

@@ -0,0 +1 @@
+Adjusts ``cmath.acosh(complex('0+nanj'))`` for recent C standards.

Modules/cmathmodule.c

@@ -1259,8 +1259,8 @@ cmath_exec(PyObject *mod)
     INIT_SPECIAL_VALUES(acosh_special_values, {
       C(INF,-P34) C(INF,-P)  C(INF,-P)  C(INF,P)  C(INF,P)  C(INF,P34) C(INF,N)
       C(INF,-P12) C(U,U)     C(U,U)     C(U,U)    C(U,U)    C(INF,P12) C(N,N)
-      C(INF,-P12) C(U,U)     C(0.,-P12) C(0.,P12) C(U,U)    C(INF,P12) C(N,N)
-      C(INF,-P12) C(U,U)     C(0.,-P12) C(0.,P12) C(U,U)    C(INF,P12) C(N,N)
+      C(INF,-P12) C(U,U)     C(0.,-P12) C(0.,P12) C(U,U)    C(INF,P12) C(N,P12)
+      C(INF,-P12) C(U,U)     C(0.,-P12) C(0.,P12) C(U,U)    C(INF,P12) C(N,P12)
       C(INF,-P12) C(U,U)     C(U,U)     C(U,U)    C(U,U)    C(INF,P12) C(N,N)
       C(INF,-P14) C(INF,-0.) C(INF,-0.) C(INF,0.) C(INF,0.) C(INF,P14) C(INF,N)
       C(INF,N)    C(N,N)     C(N,N)     C(N,N)    C(N,N)    C(INF,N)   C(N,N)