Add functionality to handle no wastewater data

[kaitejohnson]

Goal

Currently, we can run the model without using the wastewater observations, but the package will error if nothing is passed in to the ww_data arg. This should not be the case. Instead, as a temporary solution we should pass in an empty vector that will not get used by stan, and allow the unit tests forbidding the model to be run to be skipped.

Features:

• [ ] ability to pass in an empty vector/NULL for ww_data in the call to wwinference()
• [ ] handle this in pre-processing and in model fitting. I believe the best way to do this will be passing something similar to what we did when we had the two stan models in the production repo. Upstream we indicated if the model was going to ignore the wastewater, and then for that data we passed in a single value (vector of length 1), that in that case doesn't get used because the likelihood from wastewater isn't being evaluated. https://github.com/CDCgov/wastewater-informed-covid-forecasting/blob/23ce21d50c19b60a175f5ff37da35ecbe6d6136a/wweval/R/format_data_for_stan.R#L131
• [ ] adjust vignette to use this functionality rather than still passing wastewater
• [ ] warn user that wastewater isn't passed in and therefore isn't being used

@dylanhmorris my instinct is that this would be best addressed after #149, since we will be able to properly default to n_subpops=1 whenever n_sites = 0. Let me know what you think.