Adjust default priors

[null-a]

There's no good reason for the current choice of default priors. (e.g. HalfCauchy(scale=3.0) on group standard deviations.) I'm worried that the current defaults might inadvertently give the impression that brmp has similar heuristics to brms for choosing priors, which it doesn't. I think we ought to change the default priors to avoid this. (e.g. Use a standard normal for population level coefs., and a standard half normal for standard deviations perhaps.)