Implement full decoding support for patch subjects #26

Looking through decode_header in the Git source, it looks like there are several possible encodings. Currently, we only support quoted-printable UTF-8 and ignore anything else (implemented in #25.)

To support arbitrary encodings, I think we need to:

1. Identify the encoding from the text between the first =? and the next ?
2. Use the q? or b? to determine if the content until the next =? is encoded as quoted-printable or base64
3. Decode the content
4. Use ianaindex.MIME to look up the encoding
5. Use the encoding from to convert the bytes to UTF-8