pbs: fixed condition of evaluation of rdiff backup status

Initially, the condition checking rdiff backup status was right
after calling it. But then a condition about CORRUPTED_ARGS has
been inserted, making the result of $? different from initially expected.

Due to this, the cleanup of old versions was never done, making
the backup becoming too fat very quickly.

Thus clearly define variable for rdiff backup status to have expected
condition.

slapos/recipe/pbs.py

@@ -159,9 +159,11 @@ class Recipe(GenericSlapRecipe, Notify, Callback):
                     %(remote_dir)s \\
                     $BACKUP_DIR
 
+            RDIFF_BACKUP_STATUS=$?
+
             [ "$CORRUPTED_ARGS" ] && rm -f "$CORRUPTED_FILE" "$CANTFIND_FILE"
 
-            if [ ! $? -eq 0 ]; then
+            if [ ! $RDIFF_BACKUP_STATUS -eq 0 ]; then
                 # Check the backup, go to the last consistent backup, so that next
                 # run will be okay.
                 echo "Checking backup directory..."