Run CI tests with GPU hardware acceleration

[asajeffrey]

Our performance numbers are off because we are running with software rendering.

This would also catch crashes like #19519.

cc @aneeshusa @glennw @jdm

[asajeffrey]

I'm currently running ./mach test-wpt without -z locally, the results are promising so far (2 different outcomes in the first 5K tests).

[emilio]

The point is that without software rendering we don't have consistent rendering across platforms...

[asajeffrey]

I'm hoping we can find a subset of the tests that are repeatable. Can we use a .ini file to indicate which tests have to run with software rendering?

[asajeffrey]

Running the wpt CSS tests with hardware acceleration:

  ▶ TIMEOUT [expected FAIL] /css/compositing/mix-blend-mode/mix-blend-mode-animation.html
  │ 
  │ Shutting down the Constellation after generating an output file or exit flag specified
  └ Shutting down the Constellation after generating an output file or exit flag specified

  ▶ PASS [expected FAIL] /css/css-flexbox/flex-flexitem-percentage-prescation.html

  ▶ PASS [expected FAIL] /css/css-transforms/css-transforms-3d-on-anonymous-block-001.html

  ▶ PASS [expected FAIL] /css/css-transforms/transform-input-018.html

  ▶ FAIL [expected TIMEOUT] /css/css-values/vh-interpolate-vh.html
  └   → /css/css-values/vh-interpolate-vh.html f17b571b709c853a05ae9c0f417a1e422833e879
/css/css-values/reference/all-green.html 16698156583755674f3837b45ba940d68a1d3b1f
Testing f17b571b709c853a05ae9c0f417a1e422833e879 == 16698156583755674f3837b45ba940d68a1d3b1f

  ▶ PASS [expected FAIL] /css/filter-effects/filter-contrast-003.html

  ▶ TIMEOUT [expected PASS] /css/css-values/vh_not_refreshing_on_chrome.html

  ▶ PASS [expected FAIL] /_mozilla/css/border_radius_elliptical_a.html

  ▶ FAIL [expected PASS] /_mozilla/css/iframe/hide_and_show.html
  └   → /_mozilla/css/iframe/hide_and_show.html 23b755622cc81841a23cc8431a6546dfb1401c66
/_mozilla/css/iframe/hide_and_show_ref.html df80fe8f89a9c48e16251483ba2181b39b82df44
Testing 23b755622cc81841a23cc8431a6546dfb1401c66 == df80fe8f89a9c48e16251483ba2181b39b82df44

  ▶ FAIL [expected PASS] /_mozilla/css/iframe/simple_inline_width_height.html
  └   → /_mozilla/css/iframe/simple_inline_width_height.html 03a5044f78ce7b499df792938b6eb9e682e5067b
/_mozilla/css/iframe/simple_inline_width_height_ref.html 86492cf39820505827c9af3d7c990de4a575587c
Testing 03a5044f78ce7b499df792938b6eb9e682e5067b == 86492cf39820505827c9af3d7c990de4a575587c

[jdm]

All or almost all of those are known intermittent results.