Add CI workflow with lint + headless test subset across 3 OS

[mxschmitt]

Context

This addresses the gap discussed in #660: the repo has no CI. The hard part of testing pynput is input simulation, which needs real OS events and a graphical session — that's still manual. This PR lands the easy half: linting and the small subset of tests that don't need real input.

What's added

Workflow (.github/workflows/ci.yml):

• lint job — runs ruff format --check on lib/pynput and tests/ with quote-style=preserve and line-length=80. These are the same options that produce the formatting in commit eeb7eab ("Apply ruff formatting"), so the check is green on master after the test reformat below. No project-level config file is needed; the options are passed via CLI flags so this PR doesn't have to make a project-wide tooling decision.

• test job — matrixed across {ubuntu-latest, macos-latest, windows-latest} × Python {3.10, 3.11, 3.12, 3.13} (12 cells). Runs an explicit eight-test subset chosen because they don't synthesize OS input and don't wait on self.confirm():

  • keyboard_hotkey_tests: test_parse_valid, test_parse_invalid, test_activate_single, test_activate_combo — pure parser + state machine
  • mouse_controller_tests::test_buttons and keyboard_controller_tests::test_keys — verify each platform's backend exposes every Button/Key in the base enum
  • keyboard_controller_tests::test_press_invalid and test_release_invalid — exception-path tests that never reach the OS

  Linux runs under xvfb-run -a and pip-installs evdev/python-xlib so the Xorg backend can instantiate Display().

Pre-requisites also in this PR

To make the lint job green and the (now-running) tests pass, two small fixes:

1. Format tests/ with the same formatter as lib/pynput/ (commit "Apply ruff formatting to tests/ and drop u-string prefix in _info"). The May 12 ruff commit only touched lib/pynput/. Running the same ruff format --config 'format.quote-style=preserve' --config 'line-length=80' on tests/ produces the formatting in this commit. Also drops a u'...' prefix in lib/pynput/_info.py that's a Python-2 leftover and was the only remaining diff in lib/.

2. Update test callbacks for the v1.8.0 injected parameter (commit "Update test callbacks for the v1.8.0 injected parameter"). v1.8.0 added a trailing injected boolean to all listener callbacks. The arity-clipping in pynput._util._wrap protects callbacks passed directly via Listener(on_X=...), but tests/__init__.py:assert_event forwards args through a *args wrapper that bypasses _wrap, so any callback going through it must accept the new arity. Eleven callbacks across mouse_controller_tests.py and keyboard_controller_tests.py were stale; fixed mechanically by adding injected to their signatures. Confirmed locally on Windows: test_left/right/up/down previously raised TypeError: lambda takes 2 positional arguments but 3 were given and now run cleanly.

What's intentionally out of scope

• Tests that synthesize input via Controller.press/click/scroll/type/move. macOS GitHub runners can't grant Accessibility permission, so they can never run there cleanly; shared Windows runners would also flake. These remain manual, per Collaborate on testing #660.
• All listener tests — they call self.notify('Move mouse...') and self.confirm(...) and require a human at the keyboard.
• pylint. The existing pylintrc produces a -3.37/10 rating against current master out of the box; making it green requires either many code edits or many # pylint: disable= annotations, both larger scope decisions than belong in this PR.
• A project-level pyproject.toml/ruff.toml. The CLI flags inline in the workflow are deliberately minimal; if you'd prefer a config file I'm happy to add one in a follow-up.

Verification

• ruff format --check lib/pynput tests is green locally with the options above.
• The 8 headless-safe tests pass locally on Windows 11 / Python 3.13 in 8 seconds.
• The 4 previously-broken mouse_controller_tests move tests (test_left/right/up/down) now run without TypeError — outside the CI subset, but the lambda fix unblocks anyone running the suite manually.

[mxschmitt]

cc @moses-palmer some initial CI pipeline to run the existing tests for changes on the master branch.