Update Flash Attention to use processHeadsFlashAttentionOptV2

Author: orionpapadakis

src/main/java/org/beehive/gpullama3/tornadovm/layers/type/fp16/Qwen3FP16FFNLayers.java

@@ -288,7 +288,7 @@ protected TaskGraph createFFNLayerTaskGraph(int layerIndex) {
 
         // Flash Attention
         unifiedLayer.task("attention",
-                TransformerComputeKernelsLayered::processHeadsFlashAttention,
+                TransformerComputeKernelsLayered::processHeadsFlashAttentionOptV2,
                 context,
                 qwen3State.wrapQ,             // query vectors
                 qwen3State.wrapKeyCache,      // key cache