UPDATE for 04/20: More realistic vent count projections, added "slow ramp" initialization of counts, updated viz hooks

[michaelchughes]

Several big changes here (to Workflows and to core Simulation syntax)

Changes to workflows

• Added workflow fixinit_20200420

• • Updated duration distributions for most states
• • Makes much more realistic projections of time on vent that match latest data

• Updated plot functionality

• • Allows setting scenario titles on the HTML viz, so easy to track
• • Allows links to resources (like underlying summary csv files and params.txt files)
• • Has table of contents at top of HTML so you can jump quickly to the plot you want.
• • Adds anchor tags to each plot, so you can use "#n_Presenting" like modifiers to pass URLs to specific plots.

Changes to params.json and run_forecast.py/PatientTrajectory.py

• Rename parameter that provides initial counts "initnum{state}" instead of "num_{state}"

• • Have edited all workflows to fix this automatically

• Can now do "slow ramp" initialization, where params.json provides incoming counts for last Tpast days (using --num_past_timesteps kwarg).

Use syntax like this in params to say (add 15 subjects 3 days ago, 10 subjects 2 days ago, 5 subjects 1 ago, and 0 now:

"init_num_OnVentInICU":   {"-3": 15, "-2":10, "-1": 5, "0": 2},

• Can now set any initial patients to have some duration already "spent", so it looks more like steady state at beginning. Simulation will first pick a duration for the first state usual, but for initial patients will randomly decide to take off say 1 days (with 30% proba), 2 days (with 10% proba) or more. Will always leave at least 1 day in the initial state.

To specify the distribution, use syntax like below -- pmf_initial_duration_spent_{state} -- :

"pmf_initial_duration_spent_OnVentInICU": {
    "0": 0.5, "1": 0.3, "2": 0.1, "3": 0.05, "4": 0.05}