heapdict implements the MutableMapping ABC, meaning it works pretty much like a
regular Python dict. It's designed to be used as a priority queue, where items
are added and consumed as follows:

  hd = heapdict()
  hd[obj1] = priority1
  hd[obj2] = priority2
  # ...
  (obj, priority) = hd.popitem()

Compared to an ordinary dict, a heapdict has the following differences:
- popitem():
  Remove and return the (key, priority) pair with the lowest priority, instead
  of a random object.
- peekitem():
  Return the (key, priority) pair with the lowest priority, without removing it.

Unlike the Python standard library's heapq module, the heapdict supports
efficiently changing the priority of an existing object (often called "decrease-
key" in textbooks). Altering the priority is important for many algorithms such
as Dijkstra's Algorithm and A*.