implement pp_surv for varying_coef model

[jburos]

Using survivalstan.utils.plot_pp_survival with a varying-coef model will currently yield the following error:

--> 844             model['fit'].extract()[element]
    845             , columns=varnames
    846         )

KeyError: 'y_hat_time'

Because the generated quantities block does not yet compute y_hat_time.