ccflow models for database access

ccflow-db provides public, domain-neutral database callable models and publishers for ccflow workflows. It owns connection configuration, SQL execution, tabular reads, bulk writes, transaction handling, upsert/merge patterns, database-backed cache adapters, and Hydra config groups exposed through the lerna plugin entry point.

It should stay generic across domains and database backends. Domain schemas, provider-specific assumptions, and application deployment conventions belong outside this package.

• Implemented: package scaffold, version metadata, SQLiteConfig, parameterized SQLiteQueryModel, SQLiteKeyExistsModel, SQLiteTableWriteModel for append, replace, and primary-key upsert writes, byte-oriented SQLiteCacheStore for use with ccflow-etl cache format models, and the cache=sqlite config group.
• Partial: SQLite uses the Python standard library as the first local backend; broad SQLAlchemy-style engine/session management and backend extras remain future work.
• Missing: dataframe reads, chunked writes, dry-run SQL rendering, merge helpers beyond SQLite upsert, and non-SQLite integration tests.

• Depends on ccflow for callable model and publisher interfaces.
• Depends on ccflow-etl for cache format/result contracts.
• Uses sqlite3 for the first local backend and may add a broader abstraction layer such as SQLAlchemy plus optional backend extras later.
• Must not depend on finance packages or application-specific packages.

Default tests should use SQLite or in-memory database fixtures. Backend-specific tests such as Postgres should be opt-in and skipped unless the required environment variables are set.