feat(l1): pre-extract direct memory rules

[YOMXXX]

Summary

• add a conservative high-confidence rule pre-extractor for explicit persona and instruction statements
• bypass the LLM only when every new user message is covered by direct rules
• keep complex or partially covered batches on the existing LLM extraction path

Tests

• npm test -- src/core/record/l1-extractor.test.ts
• npm test
• npm run build

Refs #82

[Maxwell-Code07]

We have received multiple PRs submitted by you. Thank you for your consistent contributions. We will review them one by one and provide feedback accordingly.