Add a Vulkan fp16 variant

Lowered with eager attention and rank-3 attention (the singleton batch folded
into the head dim), mean pooling written with keepdim so the reduction stays on
the GPU, and the token embedding kept off texture storage, without which fp16
returns silently wrong values (pytorch/executorch#22333).

Verified against the CPU reference at cosine >= 0.99997 on an Adreno 840 and a
Mali-G76.

Requires the ExecuTorch 1.4.1 Vulkan runtime from the v0.10.0-libs-1.4.1
artifacts.

.gitattributes

@@ -38,3 +38,4 @@ multi-qa-MiniLM-L6-cos-v1_xnnpack.pte filter=lfs diff=lfs merge=lfs -text
 xnnpack/multi_qa_minilm_l6_cos_v1_xnnpack_fp32.pte filter=lfs diff=lfs merge=lfs -text
 coreml/multi_qa_minilm_l6_cos_v1_coreml_fp16.pte filter=lfs diff=lfs merge=lfs -text
 coreml/multi_qa_minilm_l6_cos_v1_coreml_fp32.pte filter=lfs diff=lfs merge=lfs -text
+vulkan/multi_qa_minilm_l6_cos_v1_vulkan_fp16.pte filter=lfs diff=lfs merge=lfs -text

vulkan/config.json

@@ -0,0 +1,47 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "multi_qa_minilm_l6_cos_v1",
+  "family": "sbert",
+  "capabilities": [
+    "text-embedding"
+  ],
+  "backend": "vulkan",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "multi_qa_minilm_l6_cos_v1_vulkan_fp16.pte",
+      "precision": "fp16",
+      "quantized": false,
+      "default": true,
+      "methods": {
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                510
+              ],
+              "dtype": "int64"
+            },
+            {
+              "shape": [
+                1,
+                510
+              ],
+              "dtype": "int64"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                384
+              ],
+              "dtype": "float32"
+            }
+          ]
+        }
+      }
+    }
+  ]
+}

vulkan/multi_qa_minilm_l6_cos_v1_vulkan_fp16.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9da2740169a6f3628a5f8cef092235e9d15f4918e6e9545240166a5201633505
+size 45208066