Exclude wastewater forecast-date-locs in retro that we excluded in real-time from Feb-Mar

[kaitejohnson]

Goal

We have records of which locations we used the "fall back" hospital admissions only model in real-time when we submitted from February - end of April in this repository (see example here)

To be sure that our model isn't performing worse in the retrospective than in real-time, we should apply these exclusions to the retrospective analysis

Open questions:

• should this just be for the Hub comparison? I think yes
• should we still include the full set of forecasts scores in the main analysis? I think yes
• should we subset the Hub analysis so it is Feb-end of Mar real time vs Oct- Feb fully retrospective with no exclusions?