The Distribution of the Maximum of a Gaussian Process: Rice Method Revisited