Strictly speaking, the optimizer computes the maximum of the objective function, not the maximum of the likelihood. The objective function includes the prior too. That said, when putting together a Stan model, the distinction between prior and likelihood is not clear, as each is just more data, as far as the objective function is concerned.