Investigate Array API usage in scanpy codebase

[ilan-gold]

What kind of feature would you like to request?

Additional function parameters / changed functionality / changed defaults?

Please describe your wishes

The array api is somewhat mature to the point of being using in scipy for interoperability between dense CPU/GPU functions (see https://docs.scipy.org/doc/scipy/dev/api-dev/array_api.html). It is coming with the new sparse format that may come into existence: pydata/sparse#618

So it would be good to start experimenting with this in scanpy

[c0nleyinnnn]

Hi Ilian, I'm very excited to be able to call Apple silicon's GPU acceleration on scanpy in the future. I noticed that scanpy just released an update, does that mean we're about to meet the array-api revamped scanpy? I don't know how the current development progress, are you convenient to reveal how much performance improvement has been brought by invoking GPU acceleration on MacBook? If so, I'd be more than happy to be your beta user (on my 32GB RAM M2 Pro MacBook Pro)
Finally, thank you so much for your work and that of your colleagues in contributing such wonderful tools to the single-cell omics community

[ilan-gold]

@c0nleyinnnn This issue is still open with no PR, so we have not started looking into it here, although I've begun work in anndata. As I mentioned previously, MLX (Apple's framework) is still not array-api compliant: ml-explore/mlx#48