WIP: Complete reimplementation of getindex and setindex

[meggart]

Since there were quite some performance and correctness bugs related to batchgetindex reported that proved to be quite hard to patch, I started this branch to rewrite the getindex/setindex implementation, in particular #135 #136 #137 and #138 . This makes the "strange indexing" schemes like Bool arrays, Arrays of CartesianIndex and their mixtures with integers and ranges much resolve within the normal indexing paths.

This seems to improve type stability, performance and correctness for the reported cases.

[rafaqz]

Sorry @meggart this might need a rebase after the TestTypes PR was merged

[meggart]

Thanks to everyone helping with this and for all the comments. For now I would really like to merge this, even if we delay the release of a new DiskArrays version for 1-2 weeks. It would just be much easier to test downstream packages on this and to fix unrelated bugs without running into merge conflicts. @rafaqz @felixcremer any objections?

[lazarusA]

are all compacts bounds up to date? I tried to tested it with the latest YAXArrays main and Zarr and NetCDF... and I got a lot of ERROR: Unsatisfiable requirements detected for package.

Oh.. ok... if this is merged now, I could test what it breaks 😄

[meggart]

are all compacts bounds up to date? I tried to tested it with the latest YAXArrays main and Zarr and NetCDF... and I got a lot of ERROR: Unsatisfiable requirements detected for package.

Yes because this is marked as breaking we could even tag a release after merging without breaking the downstream packages, since we would need to increase their compats anyway before any of this gets into user's julia installations.

[meggart]

And I have already created branches for NCDatasets, Zarr and NetCDF and there all tests pass so far.

[lazarusA]

don't forget YAXArrays.jl 👍🏻 , looking forward to testing them...