It allows xml parsers to understand which testcase failed the run.
Otherwise CI frameworks like jenkins wouldn't be able to classify every
single testcase. Prior to this patch testlib was using
testsuites.failures and testsuite.failures only. These are simply
reporting the number of failures.