py3: write unicode bom without using bytes() type

Since the file is opened in utf-8 mode, it needs to be written to using
unicode text, not a b'' string. In python3, '\xef\xbb\xbf' becomes ''
which is definitely not a BOM. Writing the unicode escaped codepoint
allows the codecs.open() encoding to correctly choose the right bytes
for the BOM and insert it as needed.

src/calibre/library/catalogs/csv_xml.py

@@ -107,7 +107,7 @@ class CSV_XML(CatalogPlugin):
             outfile = codecs.open(path_to_output, 'w', 'utf8')
 
             # Write a UTF-8 BOM
-            outfile.write('\xef\xbb\xbf')
+            outfile.write(u'\ufeff')
 
             # Output the field headers
             outfile.write(u'%s\n' % u','.join(fields))