Add PA_PS 8-wave kernel for MI308 with co-execution

[quintinwang5]

Motivation

Performance optimization for PA_PS on MI308 with co-execution enabled. To use this feature, just add --wave_per_tg 8

python op_tests/test_pa_ps.py -b 80 -n 10,1 -q 4 --block_size 1024 --quant_type per_1024x128 --wave_per_tg 8

Technical Details

Original 4-wave implements cannot benefit from MFMA co-execution .
The default option is wave_per_tg=4(same as before).

Test Plan

1. Correctness of gqa_raito=10,16, q_len_with_mtp=1,2,3,4 (for 8-wave, only q_len * gqa_ratio <=48 cases are supported)
2. Performance gains compared with 4-wave implements.

Test Result

1. Passed.
2. Best performance shape: batch=80, kv_seq_len=10240 ~ 16384 (Not all kernels have been well-optimized now).

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:triton-355	Run Triton tests on MI355 in addition to MI325
ci:sglang	SGLang integration tests
ci:atom	ATOM benchmark (DeepSeek-R1 + GPT-OSS)
ci:vllm	vLLM benchmark
ci:all	All of the above

Add labels via the sidebar or gh pr edit 2630 --add-label <label>