Fix CDNA bf16 ext_vector_type build with LLVM 23

Since https://github.com/llvm/llvm-project/pull/176033 (LLVM 23), AMDGPU
builtins take ext_vector_type operands, which allow no implicit conversion
between element types. The RDNA part of this was already fixed upstream
(https://github.com/ggml-org/llama.cpp/pull/29536); the CDNA paths still
use __bf16.

This is the CDNA part of https://github.com/ggml-org/llama.cpp/pull/29536,
for the live ebuild, where the full llama-cpp-llvm-23.patch no longer
applies.
--- a/ggml/src/ggml-cuda/mma.cuh
+++ b/ggml/src/ggml-cuda/mma.cuh
@@ -1430,14 +1430,14 @@
         using floatx4_t = __attribute__((ext_vector_type(4))) float;
         floatx4_t& acc_frag = reinterpret_cast<floatx4_t&>(D.x[0]);
 #if defined(CDNA4) || defined(CDNA3) || defined(CDNA2)
-        using bf16x4_t = __attribute__((ext_vector_type(4))) __bf16;
+        using bf16x4_t = __attribute__((ext_vector_type(4))) short;
         const bf16x4_t& a_frag = reinterpret_cast<const bf16x4_t&>(A.x[0]);
         const bf16x4_t& b_frag = reinterpret_cast<const bf16x4_t&>(B.x[0]);
         acc_frag = __builtin_amdgcn_mfma_f32_16x16x16bf16_1k(a_frag, b_frag, acc_frag, 0, 0, 0);
 #elif defined(CDNA1)
 #pragma unroll
         for (int i = 0; i < 2; ++i) {
-            using bf16x2_t = __attribute__((ext_vector_type(2))) __bf16;
+            using bf16x2_t = __attribute__((ext_vector_type(2))) short;
             const bf16x2_t& a_frag = reinterpret_cast<const bf16x2_t&>(A.x[i]);
             const bf16x2_t& b_frag = reinterpret_cast<const bf16x2_t&>(B.x[i]);
             acc_frag = __builtin_amdgcn_mfma_f32_16x16x8bf16(a_frag, b_frag, acc_frag, 0, 0, 0);