Add minimum-weight T-join example

[oscarhiggott]

The MWPM decoding problem is called a minimum-weight T-join in the graph theory/ combinatorial optimisation literature (the set T is the set of detection events). Therefore pymatching is an implementation of this graph theory problem, which is a subroutine of the Chinese postman problem. It would be good to explain this connection in the readme for potential users interested in combinatorial optimisation but not familiar with QEC jargon (e.g. could add an example function that uses graph theory terminology). Similarly, it might be good to add a C++ method for solving the Chinese postman problem. PyMatching is optimised for the distribution of T-join problems that arise in QEC (where the paths in the solution are typically short), however it will likely be faster than alternative available implementations even for a broader class of more generic problems since it avoids explicitly constructing the complete graph on elements of T (though it would be useful to run benchmarks to make the comparison more concrete).