No, we're not going there. That breaks the fundamental page content manipulation algorithms, and falls down for tuples not yet stored in a page (or being examined without a pointer to the page readily at hand), and has no redeeming social value anyway compared to doing it in the proven fashion.
Tom, I was also concerned regarding that, but just thought of informing you about the option. But i think it will never fall down for tuples not stored in the page. As we have the offset and the hint bits to mention whether a tuple is there or not. Only the two byte size field will move down by my suggestion. But your intuition has the most probability of success. My concern was that it would make the page of a heap different from page of a b-tree index.