update model counts

[elray1]

In section 5, we say "42 distinct models, 53 forecast dates (one per week), 54 US locations, 4 horizons, 1 target, and 23 quantiles. In the 2021-2022 season, 23 models made predictions for 22 weeks spanning from late January 2022 to late June 2022, and in the 2022-2023 season, there were 18 models"

But 42 > 23 + 18, and no doubt there was some overlap across the seasons. What's going on?

[lshandross]

I think that I used the same numbers from the flusight manuscript but forgot that the inclusion criteria we used were different. Excluding the flusight ensemble and baseline models, it should be 42 total component models, with 25 in the first season and 31 in the second (which means 14 models forecasted for both seasons).