Revert #15512 due to TC infrastructure/ failure

[gsnedders]

Reverts #15512. This should make TC infrastructure/ pass again.

cc @foolip @Hexcles

[foolip]

@gsnedders what was the failure? Can you pepper more failing expectations until it works? It did pass CI when the PR was merged.

[jgraham]

TEST-UNEXPECTED-FAIL | /infrastructure/testdriver/actions/actionsWithKeyPressed.html | TestDriver actions: actions with key pressed - assert_array_equals: lengths differ, expected 3 got 1
@http://web-platform.test:8000/infrastructure/testdriver/actions/actionsWithKeyPressed.html:64:34
Test.prototype.step@http://web-platform.test:8000/resources/testharness.js:1587:25
Test.prototype.step_func_done/<@http://web-platform.test:8000/resources/testharness.js:1627:32
promise callback*@http://web-platform.test:8000/infrastructure/testdriver/actions/actionsWithKeyPressed.html:64:6
Test.prototype.step@http://web-platform.test:8000/resources/testharness.js:1587:25
async_test@http://web-platform.test:8000/resources/testharness.js:576:22
@http://web-platform.test:8000/infrastructure/testdriver/actions/actionsWithKeyPressed.html:39:1
TEST-UNEXPECTED-OK | /infrastructure/testdriver/actions/actionsWithKeyPressed.html | expected ERROR

Also having an infrastructure test that all browsers are expected to fail seems pretty odd.

[gsnedders]

@foolip It seems to have never run the infrastructure tests prior to merging (they don't appear under Travis or TaskCluster!). Per our revert policy, I'd rather just revert the PR causing breakage on master that landed without ever running infrastructure and let the original authors figure out what to do.

[foolip]

@gsnedders merged and I'll send a new PR to see what it does right away.

Also having an infrastructure test that all browsers are expected to fail seems pretty odd.

This is because ChromeDriver isn't released as often as Chrome itself. At least I think so. In other words, things will break again when ChromeDriver is updated.

[Hexcles]

I think infrastructure tests did run in #15512 , but because of Taskcluster Checks migration, @jgraham temporarily stopped requiring Taskcluster checks so the PR slipped through...