Import do not properly handle file encodings

When importing a xar file, all contained documents are parsed using the default encoding of the server JVM, regardless of the effective encoding of the files.

This leads to incorrect documents when running the server in an UTF-8 configuration and importing the released xar (as they are encoded using ISO-8859-1, and as they effectively contain iso-8859-1 characters, which are, BTW typos coming from French).

I already corrected this pb on the current svn snapshot. I'll post the patch in a comment.

Gilles,