model: add Hy3 (hy_v3) support with MTP speculative decoding (#25395)

* model: add Hy3 (hy_v3) architecture support

Adds Tencent Hunyuan 3 (HF architecture HYV3ForCausalLM, GGUF arch
hy_v3): a MoE decoder stack with per-head Q/K RMSNorm, a sigmoid
router with expert selection bias, an always-active ungated shared
expert, and leading dense block(s) (first_k_dense_replace).

The base implementation is ported from charlie12345's fork
(https://github.com/charlie12345/ROCmFPX, src/models/hyv3.cpp),
adapted to current mainline APIs (hparams.n_layer(), build_qkv,
build_moe_ffn with fused gate_up + scale tensors, output_s).

Note: blk.N.exp_probs_b is stored without a .bias suffix for
compatibility with existing hy_v3 GGUFs produced by that fork.

Co-Authored-By: charlie12345 <charlie12345@users.noreply.github.com>
Co-authored-by: Piotr Wilkin <ilintar@gmail.com>
Assisted-by: Claude Fable 5

tests/test-jinja.cpp

@@ -1376,6 +1376,36 @@ static void test_string_methods(testing & t) {
         "bXnXna"
     );
 
+    test_template(t, "string.format() auto numbering",
+        "{{ '<{}|{}>'.format(s, 42) }}",
+        {{"s", "hello"}},
+        "<hello|42>"
+    );
+
+    test_template(t, "string.format() manual numbering",
+        "{{ '{1}-{0}-{1}'.format('a', 'b') }}",
+        json::object(),
+        "b-a-b"
+    );
+
+    test_template(t, "string.format() named fields",
+        "{{ '{name} is {age}'.format(name='Bob', age=7) }}",
+        json::object(),
+        "Bob is 7"
+    );
+
+    test_template(t, "string.format() escaped braces",
+        "{{ '{{}} {} {{x}}'.format('mid') }}",
+        json::object(),
+        "{} mid {x}"
+    );
+
+    test_template(t, "string.format() no fields",
+        "{{ 'plain'.format() }}",
+        json::object(),
+        "plain"
+    );
+
     test_template(t, "undefined|capitalize",
         "{{ arr|capitalize }}",
         json::object(),