Wrong check?

https://github.com/TimelyDataflow/differential-dataflow/blob/fbea6a8ec599c6173642bf5b40a19ea3024d537d/src/trace/implementations/spine_fueled.rs#L410

I think this is wrong: it should be the other way around.

Consider the case in which:

• through_frontier (TF) = {(1,1)}
• pending[0].upper() (UP) = {(0,2), (1,0)}

We should not move pending[0] to merging because its frontier is not dominated by TF,
and as such we should not allow it to be merged with other batches.
However, the expression evaluates to true, since (1,0) <= (1,1).

Moreover, we should also enter the loop in the case of empty TF (as it dominates every other frontier).

I think the correct check should be

(self.through_frontier.is_empty()  
  || self.pending[0].upper().iter().all(|t1| self.through_frontier.iter().any(|t2| t1.less_equal(t2))))

Start at src/trace/implementations/spine_fueled.rs:410 and read the surrounding pending-to-merging logic. Check the frontier cases described in the issue, including an empty through_frontier, and verify that the condition matches the intended dominance behavior before running the relevant test suite.