test(tools): add a pytest suite for check_patches.py

[abtonmoy]

Summary

tools/check_patches.py gates every PR on patch-set integrity, but the linter itself had no tests. A regression in it could silently start passing bad patches — the one thing it exists to prevent. This PR adds a focused test suite and gates it in CI.

Approach

Each test builds a small fixture patch-set in a temp directory, then asserts that a given check fails on exactly the violation it targets, and that a clean set passes every check. The fixtures are generated by two helpers (make_patch, write_set) so each case perturbs a single property.

Coverage

Every check has a pass case and a fail case:

Check	Fail cases exercised
series-sync	missing-from-series, extra-in-series (no backing file), duplicate entry, wrong order
numbering	gap, duplicate number, non-NNNN-*.patch name
single-surface	a patch touching two files (two diff --git)
well-formed	a patch missing a hunk; a patch missing file headers
uxr-only-switches	an added non-uxr- HasSwitch(...)
no-brand-literals	an added quoted brand string literal (plus a check that a brand token inside a // comment is allowed)

Refactor (the test seam)

The checks previously read the patch directory from a module global. This threads a patches_dir argument through the check functions and adds:

• run_checks(patches_dir) — runs all checks against a given directory and returns the Report; this is what the tests call.
• a --patches-dir CLI flag.

The default still resolves to patches/, so python tools/check_patches.py behaves exactly as before.

CI

Runs pytest tools/tests in the existing patch-integrity job.

Verification

$ python tools/check_patches.py          # default run, unchanged
OK: all patch-set checks passed

$ pytest tools/tests -q
14 passed

Closes #11