llama : add support for EmbeddingGemma 300m (#15798)

This commit add support for the EmbeddingGemma 300m. This model supports
sliding window attention (SWA) and a new swq_type is introduced to
support symmetric SWA masking.

This commit also extracts the code from the function
llama_is_masked_swa in llama-impl.h, so that the logic can be shared
by both llm_graph_input_attn_no_cache::set_input and
llama_kv_cache::set_input_kq_mask.

With this commit the EmbeddingGemma 300m model can be converted to
to GGUF and used with llama.cpp.

Once the model has been uploaded to HuggingFace it can be used like
this:
```console
./build/bin/llama-cli -hf ggml-org/embeddinggemma-300m-GGUF:Q8_0
```

src/llama-hparams.cpp

@@ -1,6 +1,7 @@
 #include "llama-hparams.h"
 
 #include "ggml.h"
+#include <cassert>
 
 void llama_hparams::set_swa_pattern(uint32_t n_pattern, bool dense_first) {
     if (dense_first) {
@@ -178,3 +179,39 @@ uint32_t llama_hparams::n_layer_kv() const {
 
     return res;
 }
+
+bool llama_hparams::is_masked_swa(llama_pos p0, llama_pos p1) const {
+    assert(p0 >= 0 && p1 >= 0);
+
+    switch (swa_type) {
+        case LLAMA_SWA_TYPE_NONE:
+            {
+            } break;
+        case LLAMA_SWA_TYPE_STANDARD:
+            {
+                if (p1 - p0 >= (int32_t) n_swa) {
+                    return true;
+                }
+            } break;
+        case LLAMA_SWA_TYPE_CHUNKED:
+            {
+                const llama_pos pos_chunk_start = (p1 / n_swa) * n_swa;
+
+                if (p0 < pos_chunk_start) {
+                    return true;
+                }
+            } break;
+        case LLAMA_SWA_TYPE_SYMMETRIC:
+            {
+                const int32_t half_n_swa = (int32_t) n_swa / 2;
+                const int32_t pos_diff = p1 - p0;
+
+                // Mask if outside the symmetric window
+                if (pos_diff < -half_n_swa || pos_diff > half_n_swa) {
+                    return true;
+                }
+            } break;
+    }
+
+    return false;
+}