I'm writing a subclass of `pathlib.Path` with custom `__fspath__`.

To prevent bad usages (users should use `os.fspath(path)` instead of `str(path)` to access the underlying path), I would like to overwrite `str`:


```
class MyPath(pathlib.PosixPath):

  def __fspath__(self) -> str:
    # Custom fspath

  def __str__(self) -> str:
    raise TypeError(
        f'Please do not use `str()` directly:'
        ' * For display: Use, f-string, `.format` or `repr`'
        ' * To access the path string: Use `os.fspath`, as per PEP 519'
    )

  def __format__(self, format_spec) -> str:
    return format(super().__str__(), format_spec)
```

However, this is breaking pathlib internal, as `str(self)` is used at many places. It would be nice if all internal `str()` calls where replaced by some `self._str()`.

I also think it would be more semantically correct if some of the `__str__` call where replaced by `__fspath__`, like: https://github.com/python/cpython/blob/b30934e9afb0af3f8e2e5f0992445be775b3c630/Lib/pathlib.py#L748