feat(openai): 极致优化 OAuth 链路并补齐性能守护

- 优化 /v1/responses 热路径，减少重复解析与不必要拷贝\n- 优化并发与 token 竞争路径并补齐运行指标\n- 补充 OpenAI/Ops 相关单元测试与回归用例\n- 新增灰度阈值守护与压测脚本，支撑发布验收

backend/internal/service/openai_gateway_service_hotpath_test.go

@@ -0,0 +1,125 @@
+package service
+
+import (
+	"net/http/httptest"
+	"testing"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/require"
+)
+
+func TestExtractOpenAIRequestMetaFromBody(t *testing.T) {
+	tests := []struct {
+		name          string
+		body          []byte
+		wantModel     string
+		wantStream    bool
+		wantPromptKey string
+	}{
+		{
+			name:          "完整字段",
+			body:          []byte(`{"model":"gpt-5","stream":true,"prompt_cache_key":" ses-1 "}`),
+			wantModel:     "gpt-5",
+			wantStream:    true,
+			wantPromptKey: "ses-1",
+		},
+		{
+			name:          "缺失可选字段",
+			body:          []byte(`{"model":"gpt-4"}`),
+			wantModel:     "gpt-4",
+			wantStream:    false,
+			wantPromptKey: "",
+		},
+		{
+			name:          "空请求体",
+			body:          nil,
+			wantModel:     "",
+			wantStream:    false,
+			wantPromptKey: "",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			model, stream, promptKey := extractOpenAIRequestMetaFromBody(tt.body)
+			require.Equal(t, tt.wantModel, model)
+			require.Equal(t, tt.wantStream, stream)
+			require.Equal(t, tt.wantPromptKey, promptKey)
+		})
+	}
+}
+
+func TestExtractOpenAIReasoningEffortFromBody(t *testing.T) {
+	tests := []struct {
+		name      string
+		body      []byte
+		model     string
+		wantNil   bool
+		wantValue string
+	}{
+		{
+			name:      "优先读取 reasoning.effort",
+			body:      []byte(`{"reasoning":{"effort":"medium"}}`),
+			model:     "gpt-5-high",
+			wantNil:   false,
+			wantValue: "medium",
+		},
+		{
+			name:      "兼容 reasoning_effort",
+			body:      []byte(`{"reasoning_effort":"x-high"}`),
+			model:     "",
+			wantNil:   false,
+			wantValue: "xhigh",
+		},
+		{
+			name:    "minimal 归一化为空",
+			body:    []byte(`{"reasoning":{"effort":"minimal"}}`),
+			model:   "gpt-5-high",
+			wantNil: true,
+		},
+		{
+			name:      "缺失字段时从模型后缀推导",
+			body:      []byte(`{"input":"hi"}`),
+			model:     "gpt-5-high",
+			wantNil:   false,
+			wantValue: "high",
+		},
+		{
+			name:    "未知后缀不返回",
+			body:    []byte(`{"input":"hi"}`),
+			model:   "gpt-5-unknown",
+			wantNil: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := extractOpenAIReasoningEffortFromBody(tt.body, tt.model)
+			if tt.wantNil {
+				require.Nil(t, got)
+				return
+			}
+			require.NotNil(t, got)
+			require.Equal(t, tt.wantValue, *got)
+		})
+	}
+}
+
+func TestGetOpenAIRequestBodyMap_UsesContextCache(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+
+	cached := map[string]any{"model": "cached-model", "stream": true}
+	c.Set(OpenAIParsedRequestBodyKey, cached)
+
+	got, err := getOpenAIRequestBodyMap(c, []byte(`{invalid-json`))
+	require.NoError(t, err)
+	require.Equal(t, cached, got)
+}
+
+func TestGetOpenAIRequestBodyMap_ParseErrorWithoutCache(t *testing.T) {
+	_, err := getOpenAIRequestBodyMap(nil, []byte(`{invalid-json`))
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "parse request")
+}