Customers can either combine a piece with more from the same set/collection (e.g. bra and knickers in lingerie or parts of a formal suit) or they freely choose parts of an outfit and a set or piece can also be shown in a different fabric, color, pattern or design. This can be divided – also visually or by gestures – into horizontal and vertical alternatives selection.
Classic mix ’n’ match or pick ’n’ mix or paper doll dress-up games for kids usually feature 3 vertical parts, i.e. either eyes, nose, mouth or head/hat, body/top, tail/bottom, but you’ll also find 2 or 4 parts and horizontal alignment.
My first intuitive design would be to copy this approach and let users switch items by vertical section or horizontal style. The example linked to in the question has just top and bottom parts.
Outfits in those sexist fashion/lifestyle magazines often consist of a standard set of components, something like main (top, dress, shirt), jewelry (earrings, necklace, ring), accessories (gloves, glasses, hat), legwear (skirt, trousers, pantyhose), footwear (socks, shoes). Since clothes are worn in multiple layers (which are not always covering lower ones fully) and their back and side views may also be decisive, that amounts to a non-trivial design task. Magazines tend to show the pieces next to (instead of on top of) each other and a picture of a mannequin or model or starlet wearing the outfit.
Another less obvious example to look at are computer games. The inventory view of classic role playing games conventionally lets the player put clothes, armor and weapons onto their character. Elsewhere, avatars can be designed and dressed even more freely, but a shopping site will probably neither have nor need flashy 3D models of all their stock items.