Fix calculation of new infected cases

[cjbayesian]

Closes #189

[cjbayesian]

Old (broken):

New:

[cjbayesian]

This is just to get the graphs' x-axes to line up vertically. Label redundant to plot title and y-label.

[PhilMiller]

There's an approximate invariant you can use to test the admissions values, no matter how they're calculated:

sum(admissions[:d]) == hospitalization_rate * market_share * (I[d]+R[d]) - initially_hospitalized

That's saying that everyone who should now be or have been hospitalized at the subject hospital was (counted as) admitted on some particular day.

I've very roughly hacked this in to test_parameters as follows and can confirm that your change to parameters.py takes it from failing to passing:

    admissions = build_admissions_df(param)
    cumulative_admissions = admissions.cumsum()
    diff = cumulative_admissions["Hospitalized"][1:-1] - (0.05*0.05 * (param.infected_v[1:-1] + param.recovered_v[1:-1]) - 100)
    assert (diff.abs() < 0.1).all()