fix(db): normalize asyncpg database URLs to sync psycopg dialect

[Stanley-blik]

Closes #44

Summary

This PR resolves issue #44 by adding automatic URL scheme normalization in flow.store.db.normalize_db_url() when creating synchronous SQLModel engines.

Problem

When FLOW_DATABASE_URL is set to an async PostgreSQL dialect URL (postgresql+asyncpg://...), SQLAlchemy's synchronous engine fallback fails to automatically check in connection objects when with get_session(engine) as s: context blocks exit. This causes connection leaks and outputs Python GC warnings:

SAWarning: The garbage collector is trying to clean up non-checked-in connection <AdaptedConnection ...>

Changes

1. Added normalize_db_url(url) to file:///home/stano/flow/src/flow/store/db.py to rewrite postgresql+asyncpg:// and postgres+asyncpg:// schemes to postgresql+psycopg://.
2. Updated make_engine() to normalize the database URL before initializing the engine.
3. Added unit tests in file:///home/stano/flow/tests/test_db.py covering SQLite, psycopg, asyncpg normalization, and engine creation calls.

Testing

• uv run ruff check passed with zero lints.
• uv run pytest passed all 47 tests (including 4 new tests in test_db.py).