attr indexation should not depend on its presence in cw_attr_cache

For optimization sake, just make sure the original rset
fetches everyting or use complete=True but relying on
cw_attr_cache will cause very obscure bugs

closes #17079907

cubicweb_elasticsearch/entities.py

@@ -118,8 +118,7 @@
             'cwuri': entity.cwuri,
         }
         for attr in self.fulltext_indexable_attributes:
-            if attr in entity.cw_attr_cache:
-                data[attr] = entity.cw_attr_cache[attr]
+            data[attr] = getattr(entity, attr)
         # TODO take a look at what's in entity.cw_relation_cache
         return data