Refactor QuerySet count() calls

[matthewcornell]

COUNT(*) statements are expensive in postgres, so we should consider refactoring QuerySet count() calls (which do COUNT under the hood). Esp. in these cases, we basically double the execution time by first calling count() and then the query itself:

• query_forecasts_for_project()
• query_scores_for_project()
• query_truth_for_project()

A better approach is to replace calling count() with simply iterating over the results, counting along the way, and terminating if the max is reached, making sure to call iterator() so that the server-side cursor (2000-rows chunks by default) is active - see [Change QuerySet calls to use iterator() #282].