Correct data for the RTCPeerConnection API

[queengooborg]

This PR corrects the data throughout the RTCPeerConnection API, which was a big mess, mostly due to copy-paste error. The fixes include the following:

• Add data regarding the prefixed constructor This one was done in Represent support for webkitRTCPeerConnection #12196 already, though the non-prefixed version number was a little off based on my testing, so this is a quick correction.
• Updates the notes regarding the default rtcpMuxPolicy value in Chrome when constructing a new instance. The notes didn't mention what the new default value is, which is slightly unhelpful. (There's only two values available, but the note update removes the need to scroll up on the page and find that out.) Cherry-picked in Update notes regarding default value of an RTCPeerConnection option #12301.
• Converts the "Promise-based version" notes into subfeatures. It is more standard to use a subfeature rather than the notes. Additionally, they were copied and pasted all across the Opera data, so it caused some major data discrepancies. (Particularly, claiming support for features in Opera that weren't supported, wrong version numbers...) Fixes Meaningless "Promise-based version." notes for RTCPeerConnection #11158.
• Corrects the data regarding promise-based versions for Chrome. After further digging in, it turns out that most of these features had the wrong version number. New data comes from the commit history (1, 2, 3).
• Corrects the Samsung Internet and Opera/Opera Android data. In the case of Samsung Internet, it appears to have been set to 6.0 most everywhere after a mirroring from the old, incorrect wiki data (Big ol' PR to populate a lot of Samsung Internet data. #1606), and Opera's data initially came from the wiki tables (Add compat data for RTCPeerConnection #1070). When the Chrome data was corrected, however (see [All] Improvements to RTCPeerConnection.json #3287), only the Chrome, Chrome Android, and WebView data was adjusted, leaving these other browsers untouched.
• Corrects the version number for createDataChannel based upon results from the mdn-bcd-collector project (v3.3.0).
• Finally, mirrors the Chrome data to both Chrome Android and WebView. Both had been left as true for many of the features, however since we have ranged values, we're able to replace the true with ≤37 for WebView, and mirror Chrome straight to Chrome Android since we know Chrome Android has supported this API back then.

There's a lot of changes made in this PR, so I would love to get it reviewed and merged ASAP!

[foolip]

I'd like to hold off with this until @ddbeck gets a chance to review #11630, since it assumes the proposed guideline.

[foolip]

Descriptions for all returns_promise entries need updating.

[foolip]

A big PR, unfortunately it's easy to find more things wrong with the data on each pass. However, I promise that with this round, if I find even more next time, I'll file follow-up issues for the rest instead.

[foolip]

I've reviewed everything again and can't spot any more issues. Thanks for sorting out this huge mess, this is much, much better now! 🥳

[foolip]

@ddbeck I think the new subfeatures for promises are release note worthy.

[ddbeck]

@foolip Good idea. In the forthcoming release notes, I've included this with the text for #11630. If you want to review today before the release, that would be helpful (but not required).

[foolip]

@ddbeck I've reviewed the release notes after the fact, looks good, thank you!