The docs for decision_tree() say at the start that the
function can fit classification, regression, and censored regression models
which is correct.
In the documenation for the argument mode, it says
Possible values for this model are "unknown", "regression", or "classification".
This is inherited from boost_tree() - which also has an engine for censored regression in the censored package. A bunch of other models do not have that though which also @inheritParams from boost_tree so we'd need to let them inherit from something else.
The models which are getting a censored regression mode are:
bag_tree()
boost_tree()
decision_tree()
rand_forest()