Well, the answer would probably be: because Windows, Mac and Linux are not primarily touch screen platforms, and multi-touch inputs is a concept that only makes any sense on touch screen platforms.
Why would engines support an input mechanism if the platform doesn't? In some cases (e.g. iOS from Mac OS) there's a common framework because the platforms have evolved their existing OS to support touch screen devices (I'm not sure if Windows and Windows Mobile have gone the same route), but that's rather a side effect than a design decision I'm sure.
As pointed out in the comments, the newest revisions of the desktop OSs are starting to embrace touch screen support, a fact of which I was unaware. But still, additional hardware is needed for everything except tablet PCs. And while Wacom input tablets are nice, they are really no substitute for touch input on the same device as the display is rendering.
So we are talking a currently miniscule proportion of the market. That's not likely to change either, until new systems start shipping with multi-touch capable input devices as standard. And since engine creators, like everyone else, only prosper by supporting targets with decent user numbers, I'd still be surprised if there were any engines which fit your criteria. I'd imagine though, if it's a popular input mechanism, demand for the hardware will grow, and existing engines will start to integrate multi-touch support. I'd be wary of any implementations that didn't have multi-touch in as a core design concept though, since it does affect the interfaces all across the UI part of an engine.