perf: cannon-rotated WASM SIMD kernels (14KB→22KB)

Updated simd-kernels-standalone.wasm with cannon-rotated
attention + FFN fast kernels from Aether.

aether-server.mjs

@@ -373,14 +373,14 @@ const server = createServer((req, res) => {
 });
 
 // ─── Main ───────────────────────────────────────────────────────────────────
-const ggufPath = '/tmp/hf_cache/buleyean-smollm2-360m-q8_0.gguf';
+const ggufPath = '/tmp/hf_cache/smollm2-360m-q8_0.gguf';
 const tokPath = '/tmp/hf_cache/tokenizer.json';
 
 async function main() {
   simd = await loadSIMD();
   if (!existsSync(ggufPath)) {
-    console.log('[Aether] Downloading Q8_0 GGUF...');
-    execSync(`python3 -c "from huggingface_hub import hf_hub_download; hf_hub_download('forkjoin-ai/buleyean-smollm2-360m', 'buleyean-smollm2-360m-q8_0.gguf', cache_dir='/tmp/hf_cache', local_dir='/tmp/hf_cache')"`, { stdio: 'inherit' });
+    console.log('[Aether] Downloading base SmolLM2-360M Q8_0...');
+    execSync(`python3 -c "from huggingface_hub import hf_hub_download; hf_hub_download('bartowski/SmolLM2-360M-Instruct-GGUF', 'SmolLM2-360M-Instruct-Q8_0.gguf', cache_dir='/tmp/hf_cache', local_dir='/tmp/hf_cache'); import shutil; shutil.move('/tmp/hf_cache/SmolLM2-360M-Instruct-Q8_0.gguf', '${ggufPath}')"`, { stdio: 'inherit' });
   }
   if (!existsSync(tokPath)) {
     console.log('[Aether] Downloading tokenizer...');

simd-kernels.wasm

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a05084c8998119797c6e80927678ce007e3285b78c6e7e8feee223ca4bb13636
-size 14553
+oid sha256:8042fe0e017a1b693fdf548c21b59a4312d615b743738a6c74bf00e570c1ae97
+size 22087