add in day of week effects to postprocessing

[mattkinsey]

we should add in a mask with a 1week scale to include the day of week effects on reporting. This is particularly important for incident hospitalizations that we report on 1 day intervals. This could be something fairly simple like fitting a holt winters model to the historical data and applying its 'seasonal' component to our outputs.