Handling of censored results within the package

[billdenney]

In some scenarios, NA values are treated as censored while in other scenarios, censored values are treated as either -Inf or Inf. I think that it would be simpler to have censored values treated as -Inf or Inf throughout the package and have NA values on input as an error.

Does that seem reasonable to you?

An example of this difference is in the quantile_cens() function.