Deleting a memory can also strip derived data from a similar, current memory #456

Found while testing #453. The cascading delete from #436 (erase() in sentient/memory/facts.py, text matching around lines 994 and 1025) finds related data (MEMORY.md/USER.md lines, journal lines, summaries, user-model insight evidence) by loose word overlap: three in five of the deleted memory's words.

When a similar memory still exists, that removes the other memory's data too. In a live test, erasing "drives a red Honda City" also removed:

• the MEMORY.md line of the current memory "sold her red Honda City and now rides a blue Ather…";
• the insight evidence pointing at that current memory (by fact id);
• message evidence of an unrelated insight.

It happens whenever a user deletes a memory that has a close sibling, not only in #453's expiry path.

Fix:

• Remove fact evidence only by fact id, never by quote.
• For text derived data (MEMORY.md/USER.md lines, journal, summaries), skip any line that also matches a remaining current memory better than the deleted one, or that names the other memory's id.
• Add tests with near-duplicate wording, such as "lives in Pune" deleted while "moved from Pune to Bengaluru" stays.

Done when: deleting one of two similar memories leaves the other's MEMORY.md line and insight evidence intact, checked in a real run.