oembed: ignore oembed results from mastodon

[mistydemeo]

This adds a check in the oembed preview for cases where the oembed is going to contain less information than the OpenGraph; in these cases, the oembed is ignored and the OpenGraph data is retained. At the moment, the only case in here is Mastodon posts, but it could be expanded in the future.

Mastodon's oembed is JavaScript based and only contains stub HTML that's intended to be filled in by client-side JavaScript that synapse doesn't execute. It doesn't contain the actual post text. The og:description, on the other hand, does contain the post text. When synapse unilaterally prefers the oembed over OpenGraph, it ends up accidentally filling in a generic and non-useful preview in place of the actual post content.

See mastodon/mastodon#34710 for context.

Pull Request Checklist

• Pull request is based on the develop branch
• Pull request includes a changelog file. The entry should:
  • Be a short description of your change which makes sense to users. "Fixed a bug that prevented receiving messages from other servers." instead of "Moved X method from EventStore to EventWorkerStore.".
  • Use markdown where necessary, mostly for code blocks.
  • End with either a period (.) or an exclamation mark (!).
  • Start with a capital letter.
  • Feel free to credit yourself, by adding a sentence "Contributed by @github_username." or "Contributed by [Your Name]." to the end of the entry.
• Code style is correct (run the linters)

[CLAassistant]

All committers have signed the CLA.

[mistydemeo]

I also considered using the provider, but unfortunately Mastodon fills in the provider with the site name and so it's different for every Mastodon instance. The only two places where we can reliably tell this is Mastodon is a) the "View on Mastodon" text, and b) the mastodon-embed class on the root <blockquote> element.

[mistydemeo]

I've added a second commit to also skip oembed text for Tumblr. Like Mastodon, Tumblr's oembed contains stub content and JavaScript to render a rich post view; in this case, the stub content is just the URL to the post, with no content. Here's what those previews look like right now:

And here's the oembed for this post: https://www.tumblr.com/oembed/1.0?url=https://www.suppermariobroth.com/post/802233242234339328/main-blog-patreon-twitter-bluesky-small

Also like Mastodon, however, the og:description contains static content that's actually useful. I think it would be better for synapse not to overwrite the useful og:description with this content. Since Tumblr has a reliable provider_name in its oembed, I've added a list of providers to skip as a constant and updated the oembed parser to also skip replacing the description if it encounters a provider in that list. Right now, the list just contains "Tumblr", but it can be updated as needed in the future.

[mistydemeo]

I missed that there's also a PR with a different implementation in #19231.