Rewrite likelihood cutoff calculator to take in a reward distribution, and output whether this is Likelihood=1 or 0. (This will require some maths that isn't immediately obvious to me). (But I think it is ultimately important to use a learned reward model rather than a fixed one, for realistic demo purposes)

[jezgillen]