feat: speed up and stabilise test suite (#676)

## ℹ️ Description
*Provide a concise summary of the changes introduced in this pull
request.*

- Link to the related issue(s): Issue #
- Describe the motivation and context for this change.

Refactors the test harness for faster and more reliable feedback: adds
deterministic time freezing for update checks, accelerates and refactors
smoke tests to run in-process, defaults pytest to xdist with durations
tracking, and adjusts CI triggers so PRs run the test matrix only once.

## 📋 Changes Summary

- add pytest-xdist + durations reporting defaults, force deterministic
locale and slow markers, and document the workflow adjustments
- run smoke tests in-process (no subprocess churn), mock update
checks/logging, and mark slow specs appropriately
- deflake update check interval tests by freezing datetime and simplify
FixedDateTime helper
- limit GitHub Actions `push` trigger to `main` so feature branches rely
on the single pull_request run

### ⚙️ Type of Change
Select the type(s) of change(s) included in this pull request:
- [ ] 🐞 Bug fix (non-breaking change which fixes an issue)
- [x] ✨ New feature (adds new functionality without breaking existing
usage)
- [ ] 💥 Breaking change (changes that might break existing user setups,
scripts, or configurations)


## ✅ Checklist
Before requesting a review, confirm the following:
- [x] I have reviewed my changes to ensure they meet the project's
standards.
- [x] I have tested my changes and ensured that all tests pass (`pdm run
test`).
- [x] I have formatted the code (`pdm run format`).
- [x] I have verified that linting passes (`pdm run lint`).
- [x] I have updated documentation where necessary.

By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice.


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Tests**
* Ensure tests run in a consistent English locale and restore prior
locale after each run
  * Mark integration scraping tests as slow for clearer categorization
* Replace subprocess-based CLI tests with an in-process runner that
returns structured results and captures combined stdout/stderr/logs;
disable update checks during smoke tests
* Freeze current time in update-check tests for deterministic assertions
* Add mock for process enumeration in web‑scraping unit tests to
stabilize macOS-specific warnings

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

docs/TESTING.md

@@ -62,6 +62,13 @@ This project uses a layered testing approach, with a focus on reliability and fa
 - **All tests in order:**
   - Run with: `pdm run test` (runs unit, then integration, then smoke)
 
+### Parallel Execution and Slow-Test Tracking
+
+- `pytest-xdist` runs every invocation with `-n auto`, so the suite is split across CPU cores automatically.
+- Pytest now reports the slowest 25 tests (`--durations=25 --durations-min=0.5`), making regressions easy to spot in CI logs.
+- Long-running scenarios are tagged with `@pytest.mark.slow` (smoke CLI checks and browser integrations). Keep them in CI, but skip locally via `pytest -m "not slow"` when you only need a quick signal.
+- Coverage commands (`pdm run test:cov`, etc.) remain compatible—`pytest-cov` merges the per-worker data transparently.
+
 ### CI Test Order
 
 - CI runs unit tests first, then integration tests, then smoke tests.
@@ -99,4 +106,4 @@ See also: `pyproject.toml` for test script definitions and `.github/workflows/bu
 ### When to Use Which
 
 - **CI:** Composite groups are preferred for CI to catch critical failures early and avoid wasting resources.
-- **Local development:** You may prefer a unified run (`pdm run test`) to see all failures at once. Both options can be provided in `pyproject.toml` for flexibility.
+- **Local development:** You may prefer a unified run (`pdm run test`) to see all failures at once. Both options can be provided in `pyproject.toml` for flexibility.