Support preshuffled layout in indexer_k_quant_and_cache / cp_gather_indexer_k_quant_cache

[1am9trash]

Motivation

In sglang PR#23562, we bump the nsa indexer's page_size to 64, which switches the indexer attention path to the _gluon_deepgemm_fp8_paged_mqa_logits_preshuffle kernel. The kernel consumes the indexer k cache in an MFMA 16×16 preshuffled layout.

Technical Details

This PR extends indexer_k_quant_and_cache and cp_gather_indexer_k_quant_cache to support writing and gathering the preshuffled layout.

We add an optional preshuffle: bool = False argument to both ops. Default is False, so existing callers are unaffected.

• preshuffle=False (default): unchanged row-major [block_size, head_dim] layout inside each paged block.
• preshuffle=True: each block's k region is written / read as a sequence of MFMA 16×16 tiles.

Also extend the testcase with -p/--preshuffle flag.

Test Plan

Test Result

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 on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
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 2879 --add-label <label>