Some users reported running out of memory when attempting to condense

diffs on large databases. This is caused by Django's default behavior

of caching retrieved objects when evaluating a queryset. It'd end up

caching all objects it'd fetch, eating up all memory.
Using iterator() bypasses this, preventing caching and giving us the

objects directly. It should prevent the memory issues people were

hitting.

Ran an upgrade successfully. Didn't have a dataset big enough to test the

memory issues, but the Django docs lead me to believe it'll solve this.