[blackwell] sm100: recompile 9 generic CUDA kernels

[Andrewxu313]

Parent: #204 — Blackwell Phase 2 kernel port
Spec: blackwell-kernel-port-v1.md § Sub-task 2
Branch: tairan/blackwell-02-02-recompile-generic
Depends on: sub-task 1

What

Populate _sm100_extensions with 9 kernels that use only generic CUDA or SM80 mma.sync (safe to recompile for -arch=sm_100a without source changes):
marlin_grouped_gemm, fp8_blockwise_ops, marlin_transform, routing generics (4×, excluding fused_gate.cu which uses WGMMA), dispatch_scatter_3d, fused_kv_norm_rope, fused_q_absorb, fused_q_split.

Acceptance

• All 9 compile with ptxas reporting sm_100a code
• Import smoke tests pass