Unfortunately, it's really not that simple. First of all, it seems to me to be closer to 50/50 between cpu and graphics card in terms of performance bottlenecks.
Second, changing the engine code to be able to support this is extremely non-trivial. It's the sort of undertaking that could sink a project - code complexity goes through the roof, debugging becomes many times more difficult, and performance gains may not even materialize.
Some things are naturally well-suited for being multi-threaded. For example, music playback happens on another thread - it just does its thing, and the main thread doesn't need to worry about it. Other things, less so - threads have to wait for each other to get to the same place, sync up, share data, etc. So even if you have say two threads doing work instead of one, chances are you're not getting anywhere near double the performance.
Finally, it's the sort of performance optimization that's not actually very useful because it doesn't do much for the worst case, which I think is what really matters. Sure, a monster PC will (potentially, if all the other problems didn't exist) reach dizzying heights - but it was already running the game fine, while a PC that was struggling won't be helped much at all. Spending time on optimizations that primarily affect the high-end doesn't seem very worthwhile.