given 其他参数，当前参数是conjugate的
Gibbs sampling 很方便
https://zh.coursera.org/learn/mcmc-bayesian-statistics/lecture/QNaSj/conditionally-conjugate-prior-example-with-normal-likelihood
