rocSHMEM (AMD) support status

Thanks for this great project!

I have a couple of questions regarding AMD GPU support.
In the [paper](https://arxiv.org/abs/2504.19442), the authors mentioned that triton-distributed supports AMD GPUs, and also highlighted the integration of OpenSHMEM primitives.

However, while exploring this repository, I noticed that rocshmem support appears to be incomplete — with many TODO comments and no actual integration visible.

I’d appreciate it if you could clarify the following:
1. What is the current status or roadmap for integrating rocshmem support into triton-distributed?
2. Which APIs or features are currently supported for AMD GPUs?
3. If I’d like to contribute or experiment with adding rocshmem support to triton-distributed, is there any guidance or recommended approach to get started?

Thank you in advance for your time and support!