This fixes how the sample analyzer handles missing_identifier errors. It was looking at the wrong line, and missing an else clause, so it was silently allowing missing_identifier errors to pass.

In addition, this fixes the sample generation so that it uses the correct filename for the output files: it previously was looking for the first line that had a filename, which was meant to indicate a non-generated line. This change adds a new Line.generated constructor for generated lines, so that they can also have the correct filename associated with them.