For every output the ARD definition has analyses for: built (its rows
are in the study ARD, made from the definition as it is now), outdated
(made from an earlier definition, or with another
programs/study_setup.R than the one there now), not built, or
error (its last
update failed). Many people may work on one study: the study ARD is
updated output by output (update_study_ard()), and a table is made
from whatever of it is there.