perf: cherry-pick PR #172 性能优化（不含 orjson）

- tiktoken 编码器改为 @lru_cache 全局缓存，避免多实例重复初始化
- 前缀匹配按长度排序，避免短前缀抢先匹配
- AuthService 更新 last_used_at 时临时关闭 expire_on_commit，减少额外 SELECT
- UsageService.record_usage_batch 改用 bulk_insert_mappings 批量插入
- 已有记录查询增加 selectinload 预加载，避免 N+1
- Gemini normalizer 长行格式化

Co-Authored-By: AAEE86 <33052466+AAEE86@users.noreply.github.com>

src/core/api_format/conversion/normalizers/gemini.py

@@ -692,10 +692,18 @@ class GeminiNormalizer(FormatNormalizer):
                     # 关闭前面的 thinking/text block（如果还开着）
                     if ss.get("thinking_block_started") and not ss.get("thinking_block_stopped"):
                         ss["thinking_block_stopped"] = True
-                        events.append(ContentBlockStopEvent(block_index=_reserve_block_index("thinking_block_index")))
+                        events.append(
+                            ContentBlockStopEvent(
+                                block_index=_reserve_block_index("thinking_block_index")
+                            )
+                        )
                     if ss.get("text_block_started") and not ss.get("text_block_stopped"):
                         ss["text_block_stopped"] = True
-                        events.append(ContentBlockStopEvent(block_index=_reserve_block_index("text_block_index")))
+                        events.append(
+                            ContentBlockStopEvent(
+                                block_index=_reserve_block_index("text_block_index")
+                            )
+                        )
 
                     name = str(func_call.get("name") or "")
                     args = func_call.get("args")
@@ -766,10 +774,14 @@ class GeminiNormalizer(FormatNormalizer):
             # 先补齐 content_block_stop（所有已开启的 block），再发送 MessageStop
             if ss.get("thinking_block_started") and not ss.get("thinking_block_stopped"):
                 ss["thinking_block_stopped"] = True
-                events.append(ContentBlockStopEvent(block_index=_reserve_block_index("thinking_block_index")))
+                events.append(
+                    ContentBlockStopEvent(block_index=_reserve_block_index("thinking_block_index"))
+                )
             if ss.get("text_block_started") and not ss.get("text_block_stopped"):
                 ss["text_block_stopped"] = True
-                events.append(ContentBlockStopEvent(block_index=_reserve_block_index("text_block_index")))
+                events.append(
+                    ContentBlockStopEvent(block_index=_reserve_block_index("text_block_index"))
+                )
             events.append(MessageStopEvent(stop_reason=stop_reason, usage=usage_info))
 
         if "error" in chunk: