Raise OverflowError when appropriate on long->float conversion. Most of

the fiddling is simply due to that no caller of PyLong_AsDouble ever
checked for failure (so that's fixing old bugs).  PyLong_AsDouble is much
faster for big inputs now too, but that's more of a happy consequence
than a design goal.

Objects/complexobject.c

@@ -522,6 +522,8 @@ complex_coerce(PyObject **pv, PyObject **pw)
 	}
 	else if (PyLong_Check(*pw)) {
 		cval.real = PyLong_AsDouble(*pw);
+		if (cval.real == -1.0 && PyErr_Occurred())
+			return -1;
 		*pw = PyComplex_FromCComplex(cval);
 		Py_INCREF(*pv);
 		return 0;