On Simplified Bayesian Modeling For Massive Geostatistical Datasets: Conjugacy And Beyond