kernels: flash attention with gpt-oss sinks + sliding window

Add flash_attention_sinks_bf16 prefill kernel that folds the per-head
attention sink into the softmax denominator (exactly as the decode sink
kernel) and supports an optional sliding-window mask matching HF gpt-oss.

Wire it through xserv-kernels (flash_attention_sinks) and use it in
GptOss prefill, replacing the post-hoc sink approximation for an exact
match against the reference math.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

crates/xserv-model/src/gpt_oss.rs

@@ -373,9 +373,8 @@ impl GptOss {
             paged_cache.append_tokens(slot, layer_idx, &k, &v, new_tokens, pos_offset);
             let (k_full, v_full) = paged_cache.gather_kv_contiguous(slot, layer_idx);
 
-            // Flash attention for prefill (sinks handled post-hoc for simplicity)
-            // TODO: integrate sinks into flash attention for exact match
-            let attn_out = flash_attention(&q, &k_full, &v_full, true);
+            // Flash attention with gpt-oss sinks + (per-layer) sliding window.
+            let attn_out = flash_attention_sinks(&q, &k_full, &v_full, &layer.sinks, layer.window_size);
 
             let attn_merged = merge_heads_gpu(&attn_out, new_tokens, num_heads, head_dim);
             let attn_proj = matmul_2d(&attn_merged, &layer.o_proj_wt);