Validating assumption on children partition count

[edmondop]

In the query stage, we take the output partition count of the first child if the plan has children, assuming all the plans have the same partition count.

While this is effectively true, if a bug is introduced in planning, this assumption will not hold anymore, and the program might start behaving in surprising ways that are difficult to troubleshoot. Rather than taking the value from the first child, the last child, or a random child, ensuring that if there are children they all have the same output partition count is a more robust approach