UTF-8 password decoding should be supported on Windows

Hi, and thank you for your work!

Currently, `get_password` and `get_credential` on Windows assume passwords to be encoded in UTF-16. However, I have encountered more than one case where the bytes object was encoded in UTF-8 (possibly by another library) and keyring would crash with a `UnicodeDecodeError`.

I think there should be a way to let the user specify a different encoding, return the bytes object directly or just fall back to UTF-8 before throwing an error. I can work on a PR if this would be a welcome change.

Thanks!

Start by locating the Windows implementations of get_password and get_credential and trace where the returned bytes are decoded as UTF-16. Reproduce the UnicodeDecodeError with UTF-8-encoded bytes, then review the proposed choices for handling alternate encodings. Done means the chosen behavior is defined and UTF-8 input no longer crashes.

Written by the indexing model from the issue text.