Properly support SVN repositories with mixed character encodings

SVNKit renders diffs with the file encoding specified in the 'file.encoding' FISHEYE_OPTS parameter, and UTF-8 by defualt. This can be problematic if a repository contains mixed file encodings because when non-UTF8 files are written to a UTF-8 file the line ending character can be clobbered and the contents of the next line can end up on the current line. This causes problems for FIshEye's diff parser because it expects the number of lines specified in the unified diff header to match the number of lines in the diff hunk.

Investigate what it will take to support mixed encoding repositories properly.