use inverse-logit transform in BYM2 model

[Connor Donegan]

The BYM2 model in stan_icar implements the model from: Morris et al "Bayesian hierarchical spatial models: Implementing the Besag York Mollié model in stan." They apply a normal prior to the logit-transformed 'mixing' parameter rho; geostan assigns a uniform prior to rho directly.

• Use their method and test it out
• Implement the model as done in the paper unless it seems worse than the current implementation

[Connor Donegan]

• didn't update for sake of keeping a hidden prior distribution diffuse, as there's no danger here due to constraint on the parameter (0, 1)