All-at-once formulation meets the Bayesian approach: A study of two prototypical linear inverse problems.