Decide on api for competing and censoring

[uriahf]

It seems natural to consider competing-event as just another status (non-event 0, event 1, competing 2), but it is unclear how to relate to censored event which is also in some cases considered as "2".

Also: what is the default competing state model and should rtichoke allow the user different state models?

Main references to consider:

• tidycmprsk
• Validation in the presence of competing risks

[uriahf]

prodlim::Hist(time, event), riskRegression::Score(), pec::cindex

• probs:
  One row (vector) for one observation, each column stands for a state.

• If only one column is provided, the probability is for the primary event. If two columns are provided than one then the first column is for the primary event and the second is for the secondary.

• reals:
  0 stands for non event or administrative-censoring, 1 for the primary event and 2 for the secondary event.

• times_to_events
  The observed time for the first event or administrative-censoring for non-events.

Scope: time to first event.