Hungarian Algorithm for Linear Sum Assignment Problem

Hungarian algorithm for linear sum assignment. Works for square and rectangular cost matrices.

Hungarian algorithm for linear sum assignment. Works for square as well as rectangular cost matrices
(thus, for the common example of matching M workers to N tasks, M can be different from N.)
The speed is compared to two other existing matlab implementations.
This implementation seems to have an edge for rectangular cost matrices; for the tested problems it runs more than 10x faster (see graphs on github: https://github.com/ondrejdee/hungarian)