Add pg_trgm candidates to entity resolution

[mikemikimike]

Summary / Problem

Closes #59. Entity resolution currently relies on exact and embedding similarity candidates, which can miss short technical identifiers and abbreviations despite a strong lexical overlap.

Changes

• Enable PostgreSQL pg_trgm in migration 0022.
• Add a GIN trigram index for AGE Node canonical names.
• Add a lexical trigram candidate channel and fuse it with vector candidates using the existing reciprocal-rank fusion helper.
• Add unit coverage for rank fusion and an integration regression test for identifier abbreviations.

Tests

• ruff check on changed files: passed (with the repository's container-executable EXE002 limitation excluded).
• Targeted unit/guard tests: 19 passed.
• Host git diff --check: passed.
• Integration regression: skipped because no ECHO_MEMORY_DATABASE_URL / PostgreSQL+AGE service was available.
• echo-memory calibrate and AUC recalibration were not run for the same unavailable database/service limitation; no AUC value is claimed.
• Pre-PR gate: prepr-20260913T013430Z-21537, Python 3.11 medium, code-related checks passed; container lacked git.

Compatibility / Known limitations

The migration requires PostgreSQL with Apache AGE and the pg_trgm extension. Existing resolution thresholds were not changed. Full database integration and calibration remain to be run in an environment with the repository's AGE/pgvector database service.

Issue link

Closes #59

[github-actions]

All contributors have signed the CLA.
_{Posted by the CLA Assistant Lite bot.}

[mikemikimike]

I have read the CLA Document and I hereby sign the CLA