prepare method
Called once, before the first score, with the decision's targets. An
engine checks here what it has to before it can score them exactly
(LiteRT-LM finds out how its version scores after a prefill, in debug
and profile builds only: a release build scores in the mode the check
picked on every engine measured), and fails with
DeciderUnsupportedException when it cannot. Nothing by default.
Implementation
Future<void> prepare(List<String> targets) async {}