The purpose is as follows:
  - Stop using `mrb_locale_from_utf8()`.
      - Because there is no corresponding `mrb_utf8_from_locale()`.
      - Because on Windows, for example, if the code page is 932 (CP932, likely ShiftJIS), it cannot be distinguished from the second byte 0x5c (\), and returns wrong results.
  - Stop using `dirname(3)`.
      - Because leading consecutive slashes are not truncated.
        For example, if `/////a/b` is given, CRuby returns `/a`, but mruby so far returns `/////a`.
      - Because the `path` argument cannot be passed in an immutable form.
  - Stop using `_splitpath()` in the Windows implementation.
      - Because there is no support for UNC paths with up to 32767 characters.
        ref. https://learn.microsoft.com/ja-jp/dotnet/standard/io/file-path-formats#unc-paths
      - Because modifying the result of paths terminated by a directory separator.
        Previously, for example, `C:/` would return `C:.` instead of `C:/`, and `a/b/` would return `a/b` instead of `a`.