Use better assertions for computedrole

[cookiecrook]

Resolves #39118

Replaces two invalid computedrole assertions which relied on an orphaned menuitem and a previously unspecified but now incorrect role synonym.

All three of the updated computedrole tests pass in WebKit and Chromium ( and Gecko soon), and this test effectively exercises the webdriver computedrole getter. Other disputed or erroneous tests that were removed here will be handled in /html-aam and /wai-aria/role

More detail in #39118

[cookiecrook]

@jcsteh looks like Gecko is failing the new "main" role test. Would you like to suggest another innocuous one for this webdriver test, since that specific role failure is already tested in /html-aam/role/roles.html? Perhaps <div role=main> or <nav>?

[jcsteh]

ComputedRole is currently pretty broken in Gecko, as it exposes our internal role strings rather than standardised ARIA roles. I've been working on this and I have a bunch of patches up for review to fix it. Once those land, this test should pass.

That said, <article>foo</article> should pass even with the current brokenness.

[cookiecrook]

Gecko fails for <nav> too, but I have not been able to get WPT running in Firefox so I'm not going to play trial and error with the Azure Pipeline. I'll add an <h2> heading as the last attempt, but if that doesn't pass, I'll leave as-is. These are all valid failures in Gecko.

[cookiecrook]

Comments crossed. I will try <article> thanks.

[cookiecrook]

Confirmed Gecko passes the <article> test. Thanks Jamie.

Okay to merge once @patrickangle or @AutomatedTester reviews.

[jgraham]

This doesn't require review from all the people marked as reviewers. In particular if the people working on a11y are happy with this change, and it doesn't also require a WebDriver spec change, we should land it without review from a WebDriver reviewer (if it does require a spec change we could still find a way to land it as a tentative test without WebDriver review, as long as there's a spec issue open).

[juliandescottes]

Given the conversation, I think it's fine to approve this, as long as @jcsteh is happy with the change as well.