This is a nice example what Haskell can do, but the sort itself is quite inefficient and it's not a true quicksort.
In this case, as in all others, the Haskell compiler would decide what is the optimal way to implement it, given whatever other constraints you've placed on the program.
You can add additional constraints that ensure the compiled code turns out to be in-place, at a cost of verbosity.
I mean, to a very large extent, this is the reason why classical imperative languages are so easy to understand and reason about. The programmers directions are much more straight forwardly executed than in examples like these hypotheticals.
Oh, I just got to your last sentence. Basically, that this would be a lot more verbose if written in such a way that it would be in place. Which is kind of the point of this criticism. Isn't it?
Don't get me wrong, the power of Haskell is not really in question. Just that example.
Sometimes, you really do want to make commands to the bare metal regarding exactly how you want the program implemented. For general use, however, you don't: compilers are (at least these days) a lot better at finding optimizations than the typical user.
It comes down to abstraction levels: pick the one that matches the problem you're working on. The FP philosophy is "speaking about specific memory cells is way too low" in most cases, like telling it you want a sorted list.
In fact, it would probably be even better (from that selfsame philosophy) to define a sort in terms of the conditions the final list would meet (as opposed to nested, recursing conditions like in the example). Something like, "return a list where every element is <= the one after it, and where every element is also in the original list, as many times as it occurs there".
If you want to see the examples of quicksort in Haskell to enforce in-place and other requirements, see this discussion here, with links:
http://www.haskell.org/haskellwiki/Introduction/Direct_Trans...
tl;dr: Yeah, Haskell is worse at letting you force the sort to be in-place, but that can be a very good thing.
http://www.haskell.org/pipermail/haskell-cafe/2009-August/06...
Twice as slow as C but without also in-place like C and extremely similar (see the sort function here):
http://en.wikibooks.org/wiki/Algorithm_Implementation/Sortin...
It's interesting to see is how very similar Haskel and C sources fundamentally become once they really solve the same problem and not the different ones.
That said, it is highly telling that none of the actual implementations of quick sort in any of these languages are that succinct. To the point that it is still a truism to "use a library" for something such as sorting. Because that isn't easy. Do not be fooled by such a quaint example.