If you want to write such a function, it should be a global function
(customization point-style) that operates on any SizedRange (maybe
also RandomAccess?).

However, I don't believe that the desire for the largest valid index
is sufficiently motivated to be something that we should add to the
standard. You also need to account for what a function that returns
the largest valid index should return when the range is empty and
therefore there *is no* valid index.