fix: preserve per-knowledge-base sparse retrieval ranks and use them in rank fusion to avoid distorted ordering across independent FTS5 indexes - #9426

修复多个知识库各自使用独立 FTS5 索引时，直接按原始 BM25 分数合并候选导致的排序失真。

不同 FTS5 索引的 BM25 分数受各自语料规模、词频和文档长度统计影响，不能直接跨索引比较。此前合并后的位置被当作稀疏检索排名传入 RRF，可能使小知识库中的高相关结果被大知识库候选压低。

Fixes #9425

Modifications / 改动点

• 为 SparseResult 增加可选的来源排名 rank。

• FTS5 检索时保留候选在各知识库独立索引中的局部排名，RRF 使用该排名计算贡献。

• 内存 BM25 回退路径仍在统一候选集上排序，并记录全局排名，保持原有语义。

• 对未提供 rank 的旧调用保留按列表顺序计算排名的兼容行为。

• 增加跨知识库稀疏检索和 RRF 融合回归测试，测试数据均为虚构内容。

• This is NOT a breaking change. / 这不是一个破坏性变更。

Screenshots or Test Results / 运行截图或测试结果

验证步骤：

.venv/bin/ruff format --check astrbot/core/knowledge_base/retrieval/sparse_retriever.py astrbot/core/knowledge_base/retrieval/rank_fusion.py tests/unit/test_sparse_retriever.py tests/unit/test_rank_fusion.py
.venv/bin/ruff check astrbot/core/knowledge_base/retrieval/sparse_retriever.py astrbot/core/knowledge_base/retrieval/rank_fusion.py tests/unit/test_sparse_retriever.py tests/unit/test_rank_fusion.py
.venv/bin/python -m pytest -q tests/unit/test_sparse_retriever.py tests/unit/test_rank_fusion.py tests/test_kb_import.py tests/unit/test_kb_document_cleanup.py tests/unit/test_kb_manager_resilience.py tests/unit/test_kb_upload_atomicity.py tests/unit/test_knowledge_base_service_contract.py

结果：

4 files already formatted
All checks passed!
43 passed, 4 warnings in 129.61s

4 条警告来自既有 aiosqlite 原子性测试在线程结束时访问已关闭的事件循环，与本次改动无关。

Checklist / 检查清单

• 😊 If there are new features added in the PR, I have discussed it with the authors through issues/emails, etc.
  / 如果 PR 中有新加入的功能，已经通过 Issue / 邮件等方式和作者讨论过。
• 👀 My changes have been well-tested, and "Verification Steps" and "Screenshots" have been provided above.
  / 我的更改经过了良好的测试，并已在上方提供了“验证步骤”和“运行截图”。
• 🤓 I have ensured that no new dependencies are introduced, OR if new dependencies are introduced, they have been added to the appropriate locations in requirements.txt and pyproject.toml.
  / 我确保没有引入新依赖库，或者引入了新依赖库的同时将其添加到 requirements.txt 和 pyproject.toml 文件相应位置。
• 😮 My changes do not introduce malicious code.
  / 我的更改没有引入恶意代码。

Summary by Sourcery

Preserve per-knowledge-base sparse retrieval ranks and use them in rank fusion to avoid distorted ordering across independent FTS5 indexes.

Bug Fixes:

• Fix cross-knowledge-base sparse retrieval ranking by using local FTS5 ranks instead of raw BM25 scores when fusing results.
• Ensure BM25 in-memory fallback path assigns and propagates global ranks consistently for rank fusion.

Enhancements:

• Extend SparseResult with an optional rank field and make rank fusion backward-compatible with callers that do not provide ranks.

Tests:

• Add unit tests to verify sparse retriever preserves per-knowledge-base FTS ranks and that rank fusion prioritizes high-relevance results from smaller knowledge bases.