Add command-line interface (closes #44)

[jedymatt]

Adds a sqlalchemyseed CLI so a database can be seeded from data files without writing Python — the first feature from the revival roadmap (#65). Closes #44.

Usage

sqlalchemyseed data.json --url sqlite:///app.db
sqlalchemyseed seeds/ --url "$DATABASE_URL"          # directory of json/yaml
sqlalchemyseed a.json b.yaml --url sqlite:///app.db  # multiple files
python -m sqlalchemyseed data.json --url sqlite:///app.db

What's included

• Console script ([project.scripts]) and python -m sqlalchemyseed via __main__.py
• One or more files and/or directories (a directory seeds every .json/.yaml/.yml, sorted)
• DB URL from --url or the DATABASE_URL env var
• --dry-run (seed in a transaction, then roll back), --seeder basic|hybrid, --ref-prefix
• CSV support via --model (CSV isn't self-describing)
• Inserts the working directory on sys.path so model paths like models.Person resolve against the user's project
• All files seed in a single transaction — commit once, or roll back as a unit on error

Implementation is a thin, dependency-free (argparse) wrapper over the existing loaders + Seeder/HybridSeeder; no changes to the core engine.

Tests & verification

• 12 new CLI tests (each format, directories, multiple paths, dry-run, hybrid, env-var URL, and error paths) — 71 tests pass total
• Driven end-to-end against a real external project: dry-run, real seed, python -m, $DATABASE_URL, and --help all behave correctly
• uv build registers the sqlalchemyseed console script in the wheel

Bumps version to 2.2.0.

🤖 Generated with Claude Code