The question "Can any given shape be created out of polypeptide chains" is rather tricky, but let's unpack it. A protein has a fold - or topology - that determines its overall shape, but can be considered as a framework on which to hang active residues.
There are 'only' 1,200-1,400 known folds, depending on the classification scheme (CATH, SCOP) which seems quite limited. However, if we take just one of those folds - the TIM barrel - and alter the loops you can get a diverse set of structures. That paper does not discuss enzymatic function, of course.
So for small molecules, it's difficult to see any reason why you can't arrange active sites any way you like. The only theoretical difficulty is that you have to fold proteins before they can be active. It might be possible that some complicated arrangement of active site residues would prevent the whole protein from folding - although that's just speculation.
More complicated might be reactions on macromolecules. For example, there are enzymes that bind to DNA, polysaccharides, and other proteins. These longer molecules are still possible for enzymes to bind to - it generally just requires a long surface. Some of the more complex actions include that performed by topoisomerases or chaperones or whatever vaults do.
By moving from tertiary to quaternary structure, it seems like enzymes can do nearly anything that's chemically possible.