Provide no-redundant-roles exception for rowgroup

[khiga8]

Relates to: https://github.com/github/accessibility/issues/5304

Follow-up to: https://github.com/github/accessibility/discussions/4921

What

This PR adds an exception so that the no-redundant-roles rule does not flag redundant use of rowgroup for:

• <thead role="rowgroup">
• <tbody role="rowgroup">

Why

There have been 3 reported instances of false positives raised by this rule. All of these relate to redundant usage of role="rowgroup". Redundant usage of role="rowgroup" is necessary in some cases to ensure that table semantics are not dropped by screen readers.

The reported instances are:

• a screen reader/browser bug where table semantics were being dropped when display: grid or display: contents is used. Example - example in primer/react
• a screen reader/browser bug where table semantics were being dropped when table header is missing. (Related: https://github.com/github/github/pull/261899#discussion_r1121141272)

It seems appropriate to not flag redundant use of role="rowgroup" for now. We can always revisit this in the future.

Why don't we turn this rule off completely?

All the reported false positives are for role="rowgroup" which we can except, so it does not seem worth turning the rule off completely.

[accessibility-bot]

👋 Hello and thanks for pinging us! You've entered our first responder queue. An accessibility first responder will review this soon.

• 💻 On PRs for our review: please provide a review environment with steps to validate, screenshots (with alt text), or videos demonstrating functionality we should be checking. This will help speed up our review and feedback cycle.
• ⚠️ If this is urgent, please visit us in #accessibility on Slack and tag the first responder(s) listed in the channel topic.

[TylerJDev]

Looks good, thanks a ton too! ✨

[TylerJDev]

Just curious, is this needed because the object is overridden when we provide our own config/mapping? 🤔

[khiga8]

Yes, that is correct!

[khiga8]

Or at least thats how I remember it.. let me double check 👀

[khiga8]

If I'm interpreting the logic correctly, it looks like the rule will use the default { nav: ['navigation']} if the consumer isn't already passing an alternate option! This seems fine to keep!