old grader matched by substring (first 40 chars of assertion text),
which silently failed when the grader rephrased assertions. new
approach uses numbered assertions for unambiguous matching.
Co-authored-by: thierrypdamiba <thierrypdamiba@users.noreply.github.com>