Enable fitting of multiple admissions catchment areas with a user supplied mapping of subpopulations

[kaitejohnson]

Goal

Under the hood, subpopulation-level incident infections are being generated. We want users to be able to 1 generate estimates of latent hospital admissions at the count catchment level desired, and/or fit the model directly to multiple count data streams. To do this, we would start by assuming that wastewater catchment areas are the same or smaller than count catchment areas, and that all wastewater catchment areas can be contained within a count catchment area. Then the user would just need to provide a mapping of count catchment area to wastewater catchment areas, and the package could fit directly to these observations and return more granular forecasts.

Requirements

• ask user to specify subpopulation -> count catchment area mapping, with the default being that all subpopulations contribute to the global count catchment area mapping (e.g. current model is special case)
• modify wwinference.stan to take in multiple count data streams indexed by the count catchment area.
• implement aggregation to count catchment area in the model
• modify likelihood to fit to multiple count data streams
• allow the user to also specify if they would like all the latent admissions for the most granular subpopulations generated (which could be useful even in the absence of data at this granularity)