bpo-36974: Make tp_call=PyVectorcall_Call work for inherited types (GH-13699)

When inheriting a heap subclass from a vectorcall class that sets
`.tp_call=PyVectorcall_Call` (as recommended in PEP 590), the subclass does
not inherit `_Py_TPFLAGS_HAVE_VECTORCALL`, and thus `PyVectorcall_Call` does
not work for it.

This attempts to solve the issue by:
* always inheriting `tp_vectorcall_offset` unless `tp_call` is overridden
  in the subclass
* inheriting _Py_TPFLAGS_HAVE_VECTORCALL for static types, unless `tp_call`
  is overridden
* making `PyVectorcall_Call` ignore `_Py_TPFLAGS_HAVE_VECTORCALL`

This means it'll be ever more important to only call `PyVectorcall_Call`
on classes that support vectorcall. In `PyVectorcall_Call`'s intended role
as `tp_call` filler, that's not a problem.

Lib/test/test_capi.py

@@ -515,9 +515,10 @@ class TestPEP590(unittest.TestCase):
 
     def test_vectorcall(self):
         # Test a bunch of different ways to call objects:
-        # 1. normal call
-        # 2. vectorcall using _PyObject_Vectorcall()
-        # 3. vectorcall using PyVectorcall_Call()
+        # 1. vectorcall using PyVectorcall_Call()
+        #   (only for objects that support vectorcall directly)
+        # 2. normal call
+        # 3. vectorcall using _PyObject_Vectorcall()
         # 4. call as bound method
         # 5. call using functools.partial
 
@@ -541,6 +542,27 @@ class TestPEP590(unittest.TestCase):
             kwnames = tuple(kwargs)
             return pyobject_vectorcall(func, args, kwnames)
 
+        for (func, args, kwargs, expected) in calls:
+            with self.subTest(str(func)):
+                if not kwargs:
+                    self.assertEqual(expected, pyvectorcall_call(func, args))
+                self.assertEqual(expected, pyvectorcall_call(func, args, kwargs))
+
+        # Add derived classes (which do not support vectorcall directly,
+        # but do support all other ways of calling).
+
+        class MethodDescriptorHeap(_testcapi.MethodDescriptorBase):
+            pass
+
+        class MethodDescriptorOverridden(_testcapi.MethodDescriptorBase):
+            def __call__(self, n):
+                return 'new'
+
+        calls += [
+            (MethodDescriptorHeap(), (0,), {}, True),
+            (MethodDescriptorOverridden(), (0,), {}, 'new'),
+        ]
+
         for (func, args, kwargs, expected) in calls:
             with self.subTest(str(func)):
                 args1 = args[1:]
@@ -549,12 +571,10 @@ class TestPEP590(unittest.TestCase):
                 if not kwargs:
                     self.assertEqual(expected, func(*args))
                     self.assertEqual(expected, pyobject_vectorcall(func, args, None))
-                    self.assertEqual(expected, pyvectorcall_call(func, args))
                     self.assertEqual(expected, meth(*args1))
                     self.assertEqual(expected, wrapped(*args))
                 self.assertEqual(expected, func(*args, **kwargs))
                 self.assertEqual(expected, vectorcall(func, args, kwargs))
-                self.assertEqual(expected, pyvectorcall_call(func, args, kwargs))
                 self.assertEqual(expected, meth(*args1, **kwargs))
                 self.assertEqual(expected, wrapped(*args, **kwargs))