A few more updates for GPU compatibility for TensorKit

[kshyatt]

Base.require_one_based_indexing returns a Bool on some versions, not n (integer).

Converting to the full CuMatrix/ROCMatrix is inefficient/ugly but dispatching on a SubArray of a Diagonal of a CuVector (or ROCVector) is also really ugly. Not sure what the best approach is here, but this works for now (we can revisit if it's really problematic).

[lkdvos]

Is it only the Diagonal giving issues?
We could also just add a specialization for Diagonal in general, which we should be able to handle in a GPU friendly way:

function project_hermitian_native!(A::Diagonal, B::Diagonal, ::Val{anti}) where {anti}
   if anti
       diagview(A) .= imag.(diagview(B)) .* im
   else
       diagview(A) .= real.(diagview(B))
   end
end

[edit] I think we have a utility function somewhere for the imaginary part, I can't remember the name though

[kshyatt]

That's also fine, I wrote this in the post lunch carb coma there's probably a better way

[lkdvos]

I guess the diagonal specialization is something we want anyways so that might make sense?

[codecov]

Codecov Report

❌ Patch coverage is 66.66667% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ixAlgebraKitAMDGPUExt/MatrixAlgebraKitAMDGPUExt.jl	0.00%	1 Missing ⚠️
...MatrixAlgebraKitCUDAExt/MatrixAlgebraKitCUDAExt.jl	0.00%	1 Missing ⚠️

Files with missing lines	Coverage Δ
src/implementations/projections.jl	96.20% <100.00%> (+0.09%)	⬆️
...ixAlgebraKitAMDGPUExt/MatrixAlgebraKitAMDGPUExt.jl	80.88% <0.00%> (-1.21%)	⬇️
...MatrixAlgebraKitCUDAExt/MatrixAlgebraKitCUDAExt.jl	56.66% <0.00%> (-0.97%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.