Feasibility of implementing estimation with random intercepts

[tylerandrewscott]

Lower level priority, just flagging for future discussion. Want to investigate possibility of implementing a different estimation approach under the hood of stm() code that allows for a simple iid random effect in the style of glmer(~0 + (1|GROUP)). Barrier I see now is most of the hlm stuff I'm familiar with in R does not to multinomial logit...