ISSUE-154: Backend semantic vector search with pgvector and artisan embeddings

[habdullhigh]

PR Title:
ISSUE-154: Backend semantic vector search with pgvector and artisan embeddings

PR Body:
Summary
This PR implements ISSUE-154 by adding semantic vector search for artisan discovery, integrating pgvector in local backend infrastructure, and adding targeted tests that map to the issue acceptance criteria.

What changed

• Added vector-capable database runtime for local backend using pgvector-enabled Postgres.
• Added embedding support and dependencies for OpenAI + pgvector.
• Added artisan embedding field and wiring for semantic search.
• Added semantic search endpoint with hybrid ranking:
  • semantic similarity
  • reputation influence
• Added embedding service for:
  • query embeddings
  • artisan profile text enrichment
  • cache-aware embedding reuse
• Added focused tests for semantic endpoint behavior and embedding text enrichment.
• Added env and backend documentation updates for semantic search configuration.
• Included a small follow-up housekeeping commit for remaining local updates.

Acceptance criteria mapping

• Vector database is running and integrated into the backend.
  • Achieved by pgvector-enabled DB image and successful migration flow.
• Profiles and reputation stats are successfully embedded.
  • Achieved by profile text builder including reputation context and covered by tests.
• Vector search returns relevant artisans for nuanced queries.
  • Achieved by semantic endpoint plus hybrid reranking, with deterministic behavior tests.

Test evidence

• Targeted tests:
  • docker-compose exec api pytest app/tests/test_semantic_search_feature.py app/tests/test_search.py -v
• Full backend tests:
  • make test
• Result:
  • 78 passed

Notes and tradeoffs

• Reputation enrichment is integrated at the text-building layer and is test-covered.
• Full on-chain reputation ingestion depth depends on current Soroban data plumbing and can be expanded in follow-up work.

Closes #154

If you want, I can also provide a shorter reviewer-friendly version (compact PR body) and a stricter engineering version (with implementation and verification details split for easier review).

[drips-wave]

@habdullhigh Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[gabito1451]

HI @habdullhigh PLEASE help look in to Ci fails