add: fitting severity and converting into epiparameter objects

[adamkucharski]

Create a how to (or maybe paper) summarising the below:

epiverse-trace/epiparameter#250

Some specific questions:

• Individual level data with known delays and missing values. How bad is the estimate if we use simple assumption (e.g. unknown = survive) - (cholera): Functionality for time varying CFR and different delays depending on outcome simulist#36

• Individual level data with different delays to outcomes. If assume delay is the same, how bad is the approximation? (Ebola): Functionality for time varying CFR and different delays depending on outcome simulist#36

• Individual level data with unknown delays but some truncation. How good are simple methods to estimate delays: Vignette on fitting parameters and converting into epiparameter objects epiparameter#250 and how accurate are subsequent predictions about number and timing of outcomes? E.g. Adding forecasting functionality cfr#14

• Incidence data with aggregated reported. Whether cases, deaths, both. How good is recovery of daily incidence and CFR calculation? https://github.com/CarmenTamayo/Applications-Epiverse-pipelines/blob/ctc-edits/EpiEstim_cfr_estimation.R

• Incidence data with known delays. How good is CFR estimation if use continuous distribution rather than discrete: Motivate use of continuous delay distributions cfr#115

[adamkucharski]

I've added a draft vignette on CFR outlining how to handle these specific scenarios:

• Individual level data with known delays and missing values for recoveries
• Individual level data with different delays to outcomes (the above scenario is effectively imposing a very long tail for delay to recovery, with some infinite, i.e. missing).

Tagging @ntncmch and @PaulC91, as this should solve problem (1) for cholera outlined here: epiverse-trace/simulist#36