prep_data_long_surv fails on multi-event data

[jburos]

Related to #41 & #42, the current functionality of prep_data_long_surv for multi-event data (and thus for competing risks) is broken.

Here is a reproducible example:

data = survivalstan.sim.sim_data_jointmodel(N=200)
# following breaks
ldf = survivalstan.prep_data_long_surv(
  data['events'], 
  event_col = 'event_value', 
  event_name = 'event_name', 
  time_col = 'time',
  sample_col = 'subject_id')

Produces output:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-62-173ddbd763bc> in <module>
----> 1 ldf = survivalstan.prep_data_long_surv(data['events'], event_col = 'event_value', event_name = 'event_name', time_col = 'time', sample_col = 'subject_id')

~/projects/survivalstan/survivalstan/survivalstan.py in prep_data_long_surv(df, time_col, event_col, sample_col, event_name)
    652                            ].drop_duplicates().copy()
    653         assert(all(df_covars.duplicated(subset=[sample_col, time_col])
--> 654                    is False))
    655         df_multi = pd.merge(df_events, df_covars, on=[sample_col, time_col],
    656                             how='outer')

TypeError: 'bool' object is not iterable