stdin/stdout frequently have bogus encodings set

Now that we are getting more widespread usage, we are getting more encoding/decoding errors.
We need to generally go through the code, and make sure that all places that pull information into bzr (either from the filesystem, or from stdin) are properly decoding it into a Unicode string, and then all places that are write it to the user are encoding it such that it at least won't cause an exception.

John has an encoding branch https://launchpad.net/people/jameinel/+branch/bzr/encoding that improves this.