format string should be unicode instead byte string

[Robert Collins]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Implicit ascii->unicode conversions are definitely bad.

We should rather, I think, teach error.py and other places that str
should be utf8-decoded, than splatter u'' on all our strings, for a
couple of reasons:

 - we have less code that processes strings, than strings.
 - its likely to be noticably expensive to construct all the static
strings we need when compiling and loading pyc files.

The latter we should fix by doing less work at startup *anyway*, but
that is then, this is now.

- -Rob
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkqks/wACgkQ42zgmrPGrq5nGwCgl+0HMSNgs2J1QzLmewBUHJ6A
SUsAn1vujN1xmILIrsLRv15AoHmeMXmI
=zElp
-----END PGP SIGNATURE-----