feat: 章节软删除、对话左滑删除，移除已读状态

- 章节：详情页增加删除按钮，软删除（is_active=False），AI 不再修改但保留供参考
- 章节：get_chapter 增加 is_active 校验，已删除章节返回 404
- 章节：AI 生成时参考同类别已删除章节摘要
- 对话：左滑显示删除，调用 hard delete API，删除前二次确认
- 对话：根布局包裹 GestureHandlerRootView 以支持 Swipeable
- 对话：移除已读/未读状态展示及相关 i18n

api/app/tasks/memoir_tasks.py

@@ -23,6 +23,7 @@ from app.features.memoir.models import (
     MemoirImage,
     MemoirState,
 )
+from app.features.memoir import repo as memoir_repo
 from app.features.user.models import User
 from app.core.dependencies import get_llm_provider
 from app.agents.state_schema import MemoirStateSchema, SlotData, default_state
@@ -713,6 +714,12 @@ def process_memoir_segments(self, user_id: str, segment_ids: List[str]):
                         )
                     narrative = combined_text
 
+                    # 已删除章节摘要供 AI 参考
+                    archived = memoir_repo.get_archived_chapter_summaries_sync(db, user_id, chapter_category)
+                    archived_summaries = "\n".join(
+                        f"- 《{title_text}》：{preview}" for title_text, preview in archived
+                    ) if archived else ""
+
                     if llm:
                         try:
                             if not chapter:
@@ -733,6 +740,7 @@ def process_memoir_segments(self, user_id: str, segment_ids: List[str]):
                                 existing_content=existing_content,
                                 user_profile=user_profile,
                                 birth_year=user_birth_year,
+                                archived_summaries=archived_summaries,
                             )
                             narrative_response = llm.invoke(narrative_prompt)
                             new_narrative = narrative_response.content.strip()
@@ -861,12 +869,18 @@ def generate_chapter_content(self, user_id: str, stage: str, new_content: str):
                     s.content for s in sorted(chapter.sections, key=lambda x: x.order_index) if (s.content or "").strip()
                 )
 
+            archived = memoir_repo.get_archived_chapter_summaries_sync(db, user_id, stage)
+            archived_summaries = "\n".join(
+                f"- 《{title_text}》：{preview}" for title_text, preview in archived
+            ) if archived else ""
+
             if llm:
                 prompt = get_narrative_prompt(
                     stage=stage,
                     slots={},
                     new_content=new_content,
                     existing_content=existing_content,
+                    archived_summaries=archived_summaries,
                 )
                 response = llm.invoke(prompt)
                 new_narrative = response.content.strip()