Consolidate packaging into pyproject.toml and adopt ruff

[eaguad1337]

Migrated from MasoniteFramework/orm#955 (original author: @circulon, left as a draft by him until the other 3.x PRs landed — this lands it last, as intended). Reimplemented against current 3.x and hardened.

What this does

• Packaging: all project metadata moves from setup.py into the pyproject.toml [project] table (PyPI name masonite-framework-orm, cleo 2 pin, masonitedev URLs, test/seeder extras, both console scripts). setup.py is deleted; python -m build — already what the publish workflow runs — is the supported build path. Verified locally: the wheel builds and contains all 110 package files plus both entry points.
• Tooling: ruff replaces black, isort, flake8 and flake8-pyproject for linting (E/W/F/I/UP) and formatting. pytest options move from pytest.ini into pyproject. pre-commit now runs ruff plus standard hygiene hooks. The makefile exposes lint / format / format-check / check, and the CI lint gate picks them up unchanged.
• Mechanical cleanup (separate commit): f-string conversions, import sorting, yield from, modern annotations in the Model.pyi stub.

Dead code found by the new linter

ruff's F811 flagged shadowed definitions that were silently dead, removed here:

• QueryBuilder.get_connection — an early stub returning connection_class (not a connection), shadowed by the real cached-connection getter.
• QueryBuilder.get_relation — referenced a nonexistent self.owner, shadowed by the real implementation.
• A duplicated mysql schema test whose first (never-executed) variant asserted malformed SQL — mismatched parentheses, duplicate constraint names and an impossible column count — removed rather than resurrected.
• Two test files imported tests.User and immediately shadowed it with a local User model.

Differences from the original draft

• Line length stays at 79 (the draft used 110): the codebase is already black-79 formatted, so adopting ruff format at 79 is near zero-churn (24 files lightly reformatted vs a full rewrite).
• E501 stays globally ignored as it was under flake8 — long SQL strings dominate the grammar modules, where the draft needed per-directory ignores anyway.
• Dropped the bespoke isort tweaks (force-sort-within-sections) to keep import ordering compatible with the existing style.

Tests

Full suite: 1058 passed; scripts/check_test_asserts.py clean; ruff check and ruff format --check pass; wheel build verified.