crowdin: Don't strip CDATA sections

* We used to remove CDATA sections from translations, which we can
  prevent by specifying the XMLParser with an option instead of just
  relying on etree.tostring()

Change-Id: Ic1ec407626b42c6a3a306e7e3b3b1eceda1387b3
diff --git a/crowdin_sync.py b/crowdin_sync.py
index 3cb4ec1..a9d848c 100755
--- a/crowdin_sync.py
+++ b/crowdin_sync.py
@@ -137,7 +137,8 @@
         XML = XML[XML.find('\n')+1:]
 
     try:
-        tree = etree.fromstring(XML)
+        parser = etree.XMLParser(strip_cdata=False)
+        tree = etree.parse(path, parser=parser).getroot()
     except etree.XMLSyntaxError as err:
         print(f'{path}: XML Error: {err.error_log}')
         filename, ext = os.path.splitext(path)