Add all output test files as regression tests

[georgephillips]

This allows us to track which files are changing when the tests are run. Tests will now fail if the output is different.