fix mixture model exercises

[hawkrobe]

• add more hints about what priors to use in malingering question? A bunch of people tried to use gaussian priors which don't make sense. Also maybe give some hints about using _.merge and friends to data wrangle the big list of participant group assignments into the output?
• Make it clearer in 1b that we expect them to actually implement the posterior predictive for the new alien