Currently we completely ignore the fail* pbench results as they are usually repeated and the correct iteration is used for comparison, but we should at least report the number of failed iterations and especially warn in case there are only failed results available (currently no comparison is done in such case!)