Find Median from Data Stream

Implement `MedianFinder` with `add_num(num)` and `find_median()` returning the median of everything added so far, as a float. Keep a max-heap of the lower half and a min-heap of the upper half. Fix an invariant — for example the lower half never differs in size from the upper by more than one — and restore it after every insert. The median is then either one root or the average of both. Push into one heap and move its root across; pushing directly into the "right" heap without the transfer is what breaks the ordering.