fix(dev): recreate postgres container and sync DATABASE_URL to localhost

[pentaxis93]

Summary

The PostgreSQL container is still running with the old Tailscale IP binding (100.125.136.73:5433) from before commit 35f6630, which fixed docker-compose.yml to default to 127.0.0.1. The container was never recreated, so the fix isn't in effect. Additionally, server/.env still points DATABASE_URL at the Tailscale IP, and the test conftest hardcodes localhost:5433.

This causes all server tests to fail with "connection refused" on localhost:5433.

Current State

• docker-compose.yml: correctly defaults POSTGRES_BIND to 127.0.0.1 (fixed in 35f6630)
• Running container: still bound to 100.125.136.73:5433 (pre-fix)
• server/.env: DATABASE_URL points to 100.125.136.73:5433
• conftest.py: hardcodes localhost:5433 (correct for intended setup)

Fix

1. Update server/.env DATABASE_URL to use localhost:5433
2. Recreate containers: docker compose up -d (picks up the 127.0.0.1 default)
3. Verify uv run pytest connects successfully

Mobile app is unaffected — the server container reaches PostgreSQL via Docker internal networking (postgres:5432), not the host port binding.

Acceptance Criteria

• PostgreSQL container bound to 127.0.0.1:5433
• server/.env DATABASE_URL uses localhost:5433
• uv run pytest can connect to the database