Infrastructure: Update localhost references for regression tests

[howard-e]

See #2636

Node v17+ deprioritised IPv4 addresses in favour of IPv6 with a change to dns.setDefaultResultOrder. This was done without a happy-eyeballs implementation which meant the loopback address used when running the regression tests resolved to ::1 rather than the IPv4 loopback, 127.0.0.1 which geckodriver expected. So this PR now explicitly uses 127.0.0.1.

Alternatively, we could have added --dns-result-order=ipv4first to the running script, or added dns.setDefaultResultOrder('ipv4first');, but would eventually remove it as the happy-eyeballs implementation is now included in v20.

Note the current environment details for the regression tests when running the node setup:

Environment details
  node: v18.16.1
  npm: 9.5.1

---

WAI Preview Link (Last built on Tue, 11 Jul 2023 21:24:38 GMT).

[css-meeting-bot]

The ARIA Authoring Practices (APG) Task Force just discussed Infrastructure: Update localhost references for regression tests.

The full IRC log of that discussion

<jugglinmike> subtopic: Infrastructure: Update localhost references for regression tests
<jugglinmike> github: https://github.com//pull/2749
<jugglinmike> Matt_King: According to howard-e's research, this will be fixed by Node.js version 20
<jugglinmike> Matt_King: ...but if we merge this, we can fix it in the current release of Node.js
<jugglinmike> howard-e: That's right!
<jugglinmike> howard-e: This is a low-risk patch and easy to revert. I think it should be fine to merge now that you've reviewed, Matt_King.
<jugglinmike> Matt_King: If something went wrong, then the workflows that are running the regression tests would fail, right?
<jugglinmike> howard-e: Yes
<jugglinmike> Matt_King: Okay; I'll merge it later today

[mcking65]

Thank you @howard-e for this insightful research. My only questions is whether we would want to reverse this change after Node 20 is released? Is there any downside to specifying the IP instead of localhost?