Add Maximum matching algorithm in experimental

[kchristin22]

This PR introduces the Maximum Matching algorithm as presented in the paper: Distributed-Memory Algorithms for Maximum Cardinality Matching in Bipartite Graphs, using GraphBLAS operations.

TODO:

• add push-pull optimization
• convert to uint32 when the data set is small enough
• create a separate vanilla version or add pragma to differentiate the function calls based on the GraphBLAS implementation used

[DrTimothyAldenDavis]

Looks great!

The github CI failure isn't your issue -- it's because the CI is using an older GraphBLAS. I have an updated CI in the v1.2 branch now, and it uses a more recent GraphBLAS and works fine.

I've also just posted a beta release of GraphBLAS 9.4.0. It has a new feature that will help to find JIT compiler errors, like the ones you encountered occasionally. If a JIT compiler error occurs, GraphBLAS returns an error code (a new one, GxB_JIT_ERROR), instead of the older strategy where I fallback to a generic kernel.

[kchristin22]

That's great! I think I currently use GraphBLAS 9.4.0.

I still have some things left to add/fix before losing the Draft notation. Briefly:

• Test on Orkut and make sure that no seg fault occurs: I think last time I attempted that, after my last commit, it still produced an error. AUTh's HPC is facing some issues since yesterday and I can't debug. Hopefully, I will be able to do so today or tomorrow.
• Lose debug comments
• Add short paper description in comments
• Rename to advanced method, as you had pointed out that there's currently no Graph in the input set (till Bipartite kind option for a graph is supported)
• Finish with the vanilla version
• Add documentation and poster

[kchristin22]

@DrTimothyAldenDavis I believe it's ready for review!
We can investigate the vxm mask thing mentioned in the e-mail before merging this though. Whatever you deem best!