[PATCH 1 of 8] use UTF-8 to encode/decode log text

[Alexis S. L. Carvalho]

Thus spake Andrey:
> @@ -60,6 +62,7 @@ class changelog(revlog):
>          """
>          if not text:
>              return (nullid, "", (0, 0), [], "", {})
> +        text = unicode(text, CHANGELOG_ENCODING)

Should we encode/decode the whole changelog text or just the user and
comment sections?

I'm not sure about the extra section (branch name should be UTF-8, but
I don't know if binary data is forbidden), but, at least for now, I
think we don't want to encode/decode the list of files.

Alexis