bpo-46417: Cleanup typeobject.c code

[vstinner]

• Add comment to recurse_down_subclasses() explaining why it's safe
  to use a borrowed reference to tp_subclasses.
• remove_all_subclasses() no longer accept NULL cases
• type_set_bases() now relies on the fact that new_bases is not NULL.
• type_dealloc_common() avoids PyErr_Fetch/PyErr_Restore if tp_bases
  is NULL.
• remove_all_subclasses() makes sure that no exception is raised.
• Don't test at runtime if tp_mro only contains types: rely on
  _PyType_CAST() assertion for that.
• _PyStaticType_Dealloc() no longer clears tp_subclasses which is
  already NULL.
• mro_hierarchy() avoids calling _PyType_GetSubclasses() if
  tp_subclasses is NULL.

Coding style:

• Use Py_NewRef().
• Add braces and move variable declarations to the first variable
  assignement.
• Rename a few variables and parameters to use better names.

https://bugs.python.org/issue46417

[vstinner]

test_concurrent_futures timed out on Address sanitizer and had to be re-run:

1 re-run test:
test_concurrent_futures
10 slowest tests:
- test_concurrent_futures: 25 min
- test_peg_generator: 9 min 46 sec
- test_tools: 4 min 45 sec
- test_asyncio: 2 7min 18 sec
- test_weakref: 1 min 26 sec
- test_venv: 1 min 24 sec
- test_subprocess: 1 min 6 sec
- test_lib2to3: 1 min 1 sec
- test_pickle: 59.4 sec
- test_gdb: 58.5 sec

[encukou]

In PyType_GetModuleByDef, the comment & assert is now out of date:

 // mro_invoke() ensures that the type MRO cannot be empty, so we don't have
// to check i < PyTuple_GET_SIZE(mro) at the first loop iteration.
assert(PyTuple_GET_SIZE(mro) >= 1);

With a for loop, the i < PyTuple_GET_SIZE(mro) check is done before each iteration.

[vstinner]

Two years later, I don't recall why I replaced the while() loop with a for() loop. You can restore the while if you prefer.