gpt-oss: WIP llama

Model loads and runs (CPU only), but PPL is much to high
(~1500 for 1st batch vs ~200 in mainline).
Is it because of SWA, because of vocab, or did I introduce a bug somewhere?

ggml/src/ggml.c

@@ -22067,6 +22067,7 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
                 case GGML_UNARY_OP_GELU_QUICK:
                 case GGML_UNARY_OP_SILU:
                 case GGML_UNARY_OP_SWIGLU:
+                case GGML_UNARY_OP_SWIGLU_OAI:
                     {
                         n_tasks = n_threads;
                     } break;