* Replace PyObject_VectorcallDict() with PyObject_Call() in
  _call_dialect(). Create a temporary tuple to pass the positional
  dialect argument.
* Use PyUnicode_DecodeUTF32() to create an Unicode string from
  Py_UCS4 characters.
* Replace PyErr_Format("%.200s", Py_TYPE(src)->tp_name) with
  PyErr_Format("%T", src).
* Replace PyUnicode_READ_CHAR() with PyUnicode_ReadChar().
* Replace PyUnicode_READ() with PyUnicode_ReadChar().
  join_append_lineterminator() now takes an object (Unicode string)
  as argument.
* Replace PyUnicode_GET_LENGTH() with PyUnicode_GetLength().
* Replace Py_XSETREF() with assignement and Py_XDECREF().
* Replace PyDict_Pop() with PyDict_DelItem().
* Use PyType_GetSlot() to get tp_clear and tp_alloc slots.
* Add some casts to (PyObject*).
* Add an explicit #include <string.h>.