Generalize the 2020 regression report to show local unexpected failures
Example usage:
```
./mach test-wpt --release --layout-2020 --log-raw /tmp/servo.log
./tests/wpt/reftests-report/gen.py /tmp/servo.log
firefox ./tests/wpt/reftests-report/report.html
```
Produces a report similar https://community-tc.services.mozilla.com/api/index/v1/task/project.servo.layout-2020-regressions-report/artifacts/public/regressions.html, but showing unexpected reftest failures. The CI-generated one shows Layout 2020 failures (expected or not) that succeed in Layout 2013.
Replace glutin by winit + surfman 0.2
<!-- Please describe your changes on the following line: -->
This PR updates surfman to 0.2, and replaces glutin with winit+surfman.
---
<!-- Thank you for contributing to Servo! Please replace each `[ ]` by `[X]` when the step is complete, and replace `___` with appropriate data: -->
- [x] `./mach build -d` does not report any errors
- [x] `./mach test-tidy` does not report any errors
- [x] These changes do not require tests because this should all be invisible
<!-- Also, please make sure that "Allow edits from maintainers" checkbox is checked, so that we can help you if you get stuck somewhere along the way.-->
<!-- Pull requests that do not address these steps are welcome, but they will require additional verification as part of the review process. -->
Add accountable-refcell as optional build time feature
<!-- Please describe your changes on the following line: -->
It can be used with the format `./mach build --features refcell_backtrace`.
---
<!-- Thank you for contributing to Servo! Please replace each `[ ]` by `[X]` when the step is complete, and replace `___` with appropriate data: -->
- [X] `./mach build -d` does not report any errors
- [X] `./mach test-tidy` does not report any errors
- [X] These changes fix#19308 (GitHub issue number if applicable)
<!-- Either: -->
- [X] These changes do not require tests because it adds a optional build-time feature only.
<!-- Also, please make sure that "Allow edits from maintainers" checkbox is checked, so that we can help you if you get stuck somewhere along the way.-->
<!-- Pull requests that do not address these steps are welcome, but they will require additional verification as part of the review process. -->
Increase parallelism on Linux WPT testing
The time taken by each chunk is uneven, with WPT-1 the longest before this change at 30 ~ 45 minutes. This reduces it to 15 ~ 20 minutes.
Surprisingly, increasing the number of processes seems to also make `test_element_in_collection` in `/webdriver/tests/execute_script/cyclic.py` **unexpectedly pass**. This happened reliably in three different runs:
https://community-tc.services.mozilla.com/tasks/S9O27WJvSa6j2PSjcRcbBA/runs/2
