Если я правильно понимаю, один метод агрегации:
select c1,
(array_agg(c2 order by c3 desc))[1] as c2,
sum(c3)
from t1
group by c1;
Другой метод будет distinct on
:
select distinct on (c2) c1, c2, sum(c3) over (partition by c1)
from t1
order by c1, c3 desc;