perf: preload V in decode split-kv shared mem and cache partial tensors

- Preload V into shared memory alongside K to eliminate per-element KV address lookups in the inner softmax/accum loop (doubles smem)
- Cache split-KV partial tensors (o_part, ml_part) with static tensors instead of per-call allocation in both decode and paged-decode paths
- Force is_causal=True in CUDA decode backend (decode is always causal)

astrai/extension/attention_backend.py

@@ -440,8 +440,7 @@ class CudaBackend(AttentionBackend):
             kv_cache.req_pool_indices,
             kv_indptr,
             kv_cache.max_len,
-            mask=attn_mask,
-            is_causal=is_causal,
+            is_causal=True,
         )
         return out.unsqueeze(1).flatten(2)