Point collision isn't really a high priority. It's on the todo list but it will work about the same as just using a really small collision poly around where you want the point to be. There wouldn't be much in the way of performance or functionality benefits.
Currently the collision engine does use a brute force approach. I want to change this (another thing on the todo list), but it's not straightforward. Using datastructures like quadtrees or just simple buckets causes more overhead for moving objects, since every time an object moves its bucket or quadtree section must be maintained. Therefore it is a careful tradeoff between slowing down the ordinary operation of games, and then more than offsetting that by speeding up collisions. I believe it's a trade-off worth making, but in the mean time there's a lot of other stuff to be working on, and I don't think many games are seriously affected by the brute force approach. If objects are far away so their bounding boxes do not overlap, the engine can reject the collision test with a simple rect test. This is so fast you really do need thousands of tests to notice a change in the framerate. So the current system really isn't that bad, it's just adequate.