Go back to QRIteration for default SVD algo on AMD

[kshyatt]

I tried using Bisection to avoid ROCm/rocm-libraries#12188 and accuracy issues with Jacobi, but this caused lots of other problems because Bisection isn't great at handling degenerate singular values. So I think it makes sense to go back to QRIteration because the above issue I only saw rarely with PEPSKit on AMDGPU, which effectively nobody is using.

[codecov]

Codecov Report

❌ Patch coverage is 33.33333% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ixAlgebraKitAMDGPUExt/MatrixAlgebraKitAMDGPUExt.jl	33.33%	2 Missing ⚠️

Files with missing lines	Coverage Δ
...ixAlgebraKitAMDGPUExt/MatrixAlgebraKitAMDGPUExt.jl	73.77% <33.33%> (ø)

🚀 New features to boost your workflow:

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