Stages like Group or Spool when spilling, manually maintain a vector of sorter file iterators instead of just using higher level interface: Sorter itself, that already have all this logic.

This leads to code duplication and complicates potential improvements like SERVER-99562. 

It also prevents this stages of using sorter features like MergeableSorter that automatically merges spills if file iterators themself start to consume a significant amount of memory.