Marshall thought of price as the dependent variable and quantity as the independent one (determined by producers and taken as given in the market day). So having Q on the x-axis was pretty correct geometry. 
It was only when the Walrasian approach (where price is the independent variable, announced by the auctioneer in the simplified version of Walras's tatonnement) came to dominate that the axes appeared "inverted".  Indeed, Pareto's diagrams correctly had the price on the x-axis (and Pantaleoni's as well, if I remember correctly). 
Why did Walras's approach won the day despite its unrealism (in Marshall at least we know who are responsible for the adjustment, the producers)? Because price adjustment can be taken to be instanteneous (with a good leap of faith), while quantity adjustment can't be instaneous even in the wildest economists' dreams. This would complicate the analysis too much for introductory textbooks.
Why did the axes remained drawn in textbooks the way Marshall drew them even if "we" (writer not included) teach students to think in terms of Walrasian price adjustment?  Call it a QWERTY case ante-litteram induced by the tremendous popularity of Marshall's textbook. 
BTW in the recent introductory textbook by Cowen and Tabarrok both ways of thinking about demand and supply are correctly presented. 
Nicola Giocoli
Sent from my BlackBerry® wireless device from WIND