Raise warning if outcomes are not {-1,1}

[lukelikelihood]

Hey,

I noticed this issue while assisting a colleague who used hBayesDM to fit a number of models for binary decision data (2-armed bandit, fictitious learning, RPL...). By accident, he used outcome data that was coded as "0" for no reward/punishment, "1" for reward. Looking through the stan files, it looks to me as if the models will only work as intended if outcomes are coded using "-1" and "1", but using "0" and "1" does also not lead to an error or warning.

It would be good if outcomes were, by default, checked to be of the required type, and if a warning (or even an error) were raised if that is not the case.

Best,
Lukas