hack_spfa似乎不work

Nobody has claimed this yet.

当我使用在以下代码所生成的数据上跑SPFA的时候，我的SPFA程序一共进行了25,000次进队+访问边的操作。（正好是边数+点数，这正好是理论最优的复杂度）

```python
with open("test-spfa-10000-cyaron", "w") as f:
g = Graph.hack_spfa(10000, weight_limit = 20)
f.write("10000 15000\n")
for e in g.iterate_edges():
f.write("%d %d %d\n" % (e.start - 1, e.end - 1, e.weight))
````

作为对比，我构造的一个一万个点三万条边的图会进行47,005,310次操作。另外加了SLF优化的代码在cyaron生成的数据上进行了312,772（~10x SPFA）次操作，而堆优化的Dijkstra进行了25,044次操作。

任何关于hack_spfa的使用技巧和insight？

这是我构造的数据：https://paste.ubuntu.com/p/BSH873sdrN/
这是我的SPFA程序：https://paste.ubuntu.com/p/dGWJHnHQws/

Start by reproducing the reported results with Graph.hack_spfa(10000, weight_limit=20), the supplied graph data, and the linked SPFA program. Compare the generated graph's operations with the reported 25,000 and 312,772 counts, then determine whether the behavior is expected or indicates a defect; done means the behavior is explained or the issue is fixed and verified.

Written by the indexing model from the issue text.