censored weibull function

From @bob-carpenter on December 4, 2014 2:48

Aki Vehtari suggested building in a function that behaves like some R survival analysis packages:

real weibull_right_censored_log(real y, real sigma, real alpha, int z) {
  if (z == 0) return weibull_log(y,alpha,sigma);
  else return weibull_ccdf_log(y,alpha,sigma);
}

The int value z is used as an indicator as to whether the event was right censored at y or whether it was an ordinary outcome.

Aki also mentioned that it's common to have a vector of censored outcomes, all censored at the same place.

Copied from original issue: stan-dev/stan#1153

Locate the existing weibull_log and weibull_ccdf_log implementations and their tests, then review how distribution functions are exposed in the Stan Math Library. Implement the requested right-censored Weibull behavior for the event indicator, and determine from nearby APIs how shared censoring points should be represented. Done means both ordinary and right-censored outcomes have the expected log-probability behavior with coverage for the supported interface.