temporarily expose raw on_task prediction in talkMoments

Adds onTask: {label, score} to to_talk_timeline_dict's output so we
can see the actual model prediction directly in the API response
while debugging mathWord showing 0/0 on a real transcript via the
hosted endpoint. Remove once the root cause is confirmed.

handler.py

@@ -101,6 +101,7 @@ class Utterance:
             'wordCount': self.word_count,
             'numMathTerms': self.num_math_terms,
             'mathTerms': self.math_terms,
+            'onTask': self.on_task,  # temporary: raw {label, score} for debugging mathWord source
             'vocabularyTerms': self.vocabulary_terms,
             'vocabularyMatches': self.vocabulary_matches
         }