[CUDA] support sorting complex numbers

[Lyxot]

Proposed changes

Add complex64 sort, argsort, partition, and argpartition support to the CUDA backend

Summary

Remove the complex64_t exclusion from the CUDA merge sort kernels. The current merge sort is comparison-based and already supports complex64. The guard was stale.

Checklist

Put an x in the boxes that apply.

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works

[Copilot]

Pull request overview

Adds CUDA backend support for sorting complex64 inputs by removing the previous complex64 exclusion in the CUDA merge-sort implementation and extending NaN-handling for complex types.

Changes:

• Enable CUDA merge-sort kernels for complex64 by removing the complex64 guard in both single-block and multi-block paths.
• Add NaN initialization and NaN-aware comparisons for complex64 within CUDA sort comparator logic.
• Extend Python sort/argsort test coverage to include complex64 and add a complex NaN sort assertion.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
python/tests/test_ops.py	Expands sort/argsort tests to include complex64 and adds a complex NaN sort check.
mlx/backend/cuda/sort.cu	Removes the complex64 exclusion and adds complex-aware NaN init/comparison support so complex64 can be sorted on CUDA.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Lyxot]

sry I forgot to check Metal's complex support. I will fix it later.

[Lyxot]

fixed

[zcbenz]

Looks good to me, thanks!