kv-cache : fix SWA checks + disable cacheless iSWA (#15811)

ggml-ci

src/llama-hparams.h

@@ -229,7 +229,10 @@ struct llama_hparams {
     // number of layers for which has_kv() returns true
     uint32_t n_layer_kv() const;
 
-    bool is_masked_swa(llama_pos p0, llama_pos p1) const;
+    // note that this function uses different SWA parameters from those in the hparams
+    // TODO: think of a better place for this function
+    // TODO: pack the SWA params in a struct?
+    static bool is_masked_swa(uint32_t n_swa, llama_swa_type swa_type, llama_pos p0, llama_pos p1);
 };
 
 static_assert(std::is_trivially_copyable<llama_hparams>::value, "llama_hparams must be trivially copyable");