Several tests perform document normalization and then check the resulting 
tree.  Since normalization exceptions are not reported, test failures due to 
unrelated normalization issues are difficult to diagnose.

All L3 tests that perform normalization and did not capture normalization 
errors have been modified to capture and assert that there were no errors 
during normalization.  These modifications uncovered that several tests were 
failing since the parse was not namespace aware which would cause a fatal 
error when performing namespace normalization.

assertLowerSeverity was added to the test markup language to simplify 
asserting that no errors occurred during normalization.