Modernize packaging: setup.cfg to pyproject.toml + wads uv CI stub

[thorwhalen]

Modernizes replize from legacy setuptools packaging + the pre-uv CI template
to pyproject.toml (Hatchling) + the wads uv reusable-workflow stub. Part of the
fleet-wide wads repo-improvement rollout.

⚠️ Merging this PR publishes a release. With the uv CI in place, a push to
master runs validation and then the publish job, which auto-bumps the version
and uploads to PyPI. [project].version is set to 0.1.5 — exactly what is
on PyPI today — so the first post-merge publish produces 0.1.6. Nothing is
published by this PR itself: the branch run's Publish job is skipped
(publishing is gated on the default branch).

What changed

1. setup.cfg / setup.py → pyproject.toml (wads-migrate setup-to-pyproject,
then hand-corrected for the tool's known rough edges).

Full inventory carried across from setup.cfg — nothing dropped:

setup.cfg	pyproject.toml
name = replize	[project].name
version = 0.1.5	[project].version (matches the released PyPI version)
description	[project].description
long_description = file:README.md + text/markdown	readme = "README.md"
license = apache-2.0	license = "Apache-2.0" (SPDX string, not the deprecated [project.license] table); LICENSE kept
url	[project.urls].Homepage
keywords = REPL	[project].keywords (broadened: REPL, CLI, shell, subprocess, console)
console_scripts: replize = replize.replize:_replize_cli	[project.scripts]
install_requires (empty)	dependencies = []
packages = find:	Hatchling auto-discovery of replize/
include_package_data = True	no-op — the package ships no non-Python data (no MANIFEST.in, no requirements.txt in the repo)
zip_safe = False	obsolete under Hatchling/wheels

Added while here: requires-python = ">=3.10", trove classifiers, author, and the
ecosystem-standard [tool.ruff] block (without it the repo falls through to ruff's
moving default and can go red on unrelated style drift).

Verified by building: uv build produces a clean sdist + wheel, and the wheel's
entry_points.txt contains the replize console script.

2. wads CI config corrections (the migration tool's documented rough edges):

• [tool.wads.ci].project_name was emitted empty → set to "replize" (it is the
  lint/coverage target in the reusable workflow).
• testpaths was emitted as ["tests"] → set to ["replize"]. This matters:
  wads CI runs pytest --doctest-modules with no path argument, so collection
  is driven entirely by testpaths, and this repo has no top-level tests/
  directory — ["tests"] would have collected nothing while still reporting green.
  Verified locally: pytest --doctest-modules -q (no path arg, i.e. exactly what
  CI runs) collects and passes 1 doctest.
• docsrc added to [tool.wads.ci.testing].exclude_paths.
• No test extra existed to wire into [tool.wads.ci.install] (the package has no
  dependencies at all), so nothing to do there.

3. Legacy CI → uv stub (wads-migrate ci-to-uv, then ci-to-stub).

The old workflow was the pre-uv generation (actions/checkout@v2,
actions/setup-python@v2, axblack + pylint + isee, hand-rolled twine publish).
It is replaced by the 5-line stub calling
i2mint/wads/.github/workflows/uv-ci.yml@master, configured entirely from
[tool.wads.ci.*].

Secrets: the migration commands scan the old workflow for secrets.* references.
The only ones present were PYPI_USERNAME and PYPI_PASSWORD. The uv CI uses
token-only PyPI auth, so PYPI_PASSWORD (already in the stub's pass-through
block) is sufficient and PYPI_USERNAME is no longer referenced. No secret was
lost.

4. Housekeeping: added .editorconfig (wads templates ship one; this repo
predates that). Repo settings were also aligned outside this PR: homepage set to
the docs URL and topics populated from the pyproject keywords. Pages
(gh-pages / root) and Discussions were already correct.

Verification

Check	Result
Local gate test-dependents replize — baseline	1 suites: 1 pass, 0 fail, 0 no-tests
Local gate test-dependents replize — final	1 suites: 1 pass, 0 fail, 0 no-tests (unchanged)
pytest --doctest-modules -q (no path arg, as CI runs it)	1 passed
uv build	sdist + wheel OK; console script present in wheel metadata
ruff check replize	All checks passed
Branch CI	✅ success — https://github.com/i2mint/replize/actions/runs/30859035968

Branch CI jobs: Read Configuration success · Validation (3.10) success ·
Validation (3.12) success · Windows Tests success · Publish skipped ·
Publish GitHub Pages skipped.

wads.repo_audit findings: HIGH 1 → 0, MEDIUM 3 → 1, LOW 3 → 1
(after the repo-settings alignment). Cleared: no pyproject.toml, legacy
packaging files present, CI is the legacy generation, missing .editorconfig,
homepage/topics drift. No new findings introduced.

Not done here (deliberately)

• No test suite. The only doctest that exercises behaviour is marked +SKIP
  (it would open an interactive REPL), so real coverage is effectively nil. Out
  of scope for a packaging pass — filed separately as No behavioural test coverage: the only doctest is +SKIP #1
  (not closed by this PR).
• No source reformatting. The code still uses the old axblack single-quote
  style; ruff format will normalise it on the first publish push-back. Keeping
  it out of this PR keeps the diff reviewable.
• No latent bugs surfaced — the package is two small modules with no third-party
  dependencies, and it imports and runs clean on modern Python.