Update to SuiteSparse:GraphBLAS 9.1.0

[eriknw]

No description provided.

[eriknw]

@alugowski @michelp I made some progress updating python-suitesparse-graphblas to 9.1.0 (yes I know 9.2.0 was just released).

The wheel builds are failing (hooray for tests!), and I'm wondering if either of you could help take a look (I'm busy traveling this week and next week).

Changes:

• It looks like "GraphBLAS.h" moved from "include/" to "include/suitesparse", and we need to update the wheel building scripts to look here
• It looks like add_arm_to_libomp_dylib.sh script may no longer be necessary
• SuiteSparse:GraphBLAS updated how it builds (maybe in 8.3?), and we need to update too. Maybe this means using cmake instead of make as was done here: 9.1.0 conda-forge/graphblas-feedstock#71

[alugowski]

I can take a look this weekend.

To get rid of the macos OMP hack we'll need to use the new runners as per actions/runner-images#9254
We may need to keep the current one for x86, too, since we get omp from brew and AFAIK it'll still only install the native arch. That's ok, I purposely kept the two macos arches separate in anticipation of this :)

[alugowski]

FYI I've had good experience using scikit-build-core for cmake-based projects. Better try fixing what we have first, but might be worth considering for any future refactor.

[alugowski]

Would it be possible to make this into a branch on this repo so I'm able to add commits?

[DrTimothyAldenDavis]

The latest updates to the cmake build system for GraphBLAS (and all of SuiteSparse) is very stable now. So any changes you need to make would also work for a long time, for future updates to my codes.

[eriknw]

Would it be possible to make this into a branch on this repo so I'm able to add commits?

Done. update9.1.0 branch

[alugowski]

Would it be possible to make this into a branch on this repo so I'm able to add commits?

Done. update9.1.0 branch

Great! I pushed changes to that branch that appear to work. Sorry I don't think it's possible to update the PR to be from that branch instead of the one from your fork, so I opened a new PR #128 Let's move discussion there.