Add competing_risk to MonteCarloGFormula

[pzivich]

Similar to #73 it would be nice to natively handle competing events. Option would be add_competing_risk() function to add another model for a competing risk.

Would need to add evaluation in the background. Basically, would only continue Monte-Carlo procedure on non-event, non-competing risk, and non-censored rows. Shouldn't be terrible to implement...

[pzivich]

There are two options; multinomial model to predict the distinct categories, or multiple binary models (with some order)

Multinomial logit models are looking to be my best bet. Assumes that outcomes are mutually exclusive, which they should be for competing risk anyways. The downside is that I limit some of the flexibility of the g-formula. However, I am unsure how to get competing risks working well without a multinomial.

Decision is now to automate the process in the background (auto-detect categories of Y when given the data) or have the user specifically request a competing risks model. Personally, I like the auto-detection

[pzivich]

I think I am going to go with the additional pooled logit model, rather than the multinomial. This approach allows for a unique model to be specified for the competing risk. Otherwise, the multinomial model needs to take terms from both the outcome and the competing risk. This would allow greater flexibility for the user