[Fix][Kernel] Fix fragment ownership in test_mamba.py

[guangxing83]

Fix incorrect SSDChunkScanFwd results on MACA 64-lane warps by replacing four independent 32x32 loops with one 64x64 loop, ensuring consistent fragment ownership. This preserves the existing micro-block factorization without adding global-to-shared or global-to-fragment copies. Using a smaller tile also passes the tests but autotuning may still select an incorrect 64x64 configuration.

[gemini-code-assist]

Code Review

This pull request refactors the micro-block processing in tileops/kernels/mamba/ssd_chunk_scan.py by consolidating four separate parallel loops over M x M blocks into a single 2 * M x 2 * M parallel loop. This change ensures that a single full fragment domain is maintained, keeping ownership mapping consistent across both 32-lane and 64-lane GPU backends. Additionally, the variable anchor was renamed to micro_anchor. There are no review comments, and I have no additional feedback to provide.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.