McMasterPandemic: Pandemic Model

An attempt to document what is currently going wrong with sim_funs. Interaction betweenmake_state()`, testify machinery, eigenvector machinery for generating initial conditions ... We switched two defaults: (1) start from eigenvector, (2) use testify machinery if the current parameters have a testing-rate parameter >0.

I think the fundamental design problems are that

we have a tall stack of functions (calibrate_comb() → calibrate() → run_sim_loglin() → run_sim() → do_step()) and we are trying to hard to let users call any of those functions and have them just "do what you mean", i.e. flexible/clever defaults
there is a bit of recursion going on: make_ratemat() → make_state() (default: may automatically try to compute eigvecs and/or testify) → make_ratemat() (if eigvecs desired, to set up the exponential simulation)

What happens in make_state(params=pp)?

figure out testify (params specified and has testing>0?)
set base state names (no testing/age etc.)
expand_stateval_testing(state, method="untested")
call get_evec(params, testify=testify)
- call rExp()
  - call make_state(N=1,E0=1e-5,use_eigvec=FALSE, testify=FALSE)
  - call make_ratemat(state=state) (no testify)
  - testify M
  - expand_stateval_testing()
  - run_sim_range(params, state, M)
assign evec values back
sort out starting/ending dates etc. (harmless)
if state not specified, call make_state with testify=FALSE
- call get_evec()
- call rExp()
  - call make_state(N=1,E0=1e-5, use_eigvec=FALSE)
  - call `make_ratemat(
call make_ratemat
- this uses state, so will call make_state(..., params=params) [except for infection terms, which are determined by update_foi, make_state() only needs to know about the dimensions and names of the state vector: could simplify?]