[ShiTan-677]

Query results can swap [我] and [对方] when the account owner's Name2Id.rowid differs between message shards. The query layer previously detected one numeric sender ID and reused it across databases, even though those row IDs are database-local.

Resolve sender wxids through each shard's own Name2Id join in read, search (including the compressed-message fallback), recent and summary. Formatting compares the resolved wxid with the account owner and labels unresolved senders [未知] instead of guessing from a numeric ID. System-message labels are unchanged.

Validation on macOS / Python 3.12:

• The synthetic swapped-row-ID regression failed for all four query paths before the fix.
• python -m pytest -q: 90 passed, 1 skipped, including compressed-message search and unresolved-sender cases.
• Compilation, macOS shell syntax checks and loading all six MCP tools passed.

All fixtures are synthetic; no WeChat data or keys are included. The fork CI matrix passed on macOS and Windows with Python 3.10 and 3.13: run 38035160551. The upstream fork-PR workflow still requires maintainer approval.

[freanocank]

Thanks for the fix and regression tests. Adopted through #15 with your original commits preserved. The cross-shard sender regressions pass in the combined branch.

The combined suite passes all four macOS/Windows × Python 3.10/3.13 CI jobs; Windows reports 209 passed, and macOS reports 208 passed with the Windows-only test skipped.