Replace evaluate_chu with evalcast::evaluate_covid_predictions

[nmdefries]

National data is now available in covidcast so replace custom evaluate_chu with evalcast's evaluate_covid_predictions, already used for scoring state forecasts.

[brookslogan]

Have you already verified that the new national truth data match the corresponding covidHubUtils ones? (I think national hospitalizations should match except for some things that don't matter: covidcast including time_values on or before 2020-07-27 (which aren't visualized anyway), and differences from timing of updates (which don't matter as eventually both sources will receive the same updates).)

[dajmcdon]

Right. I was going to say the same thing as @brookslogan . Before we change, we've got to make sure that the data matches.