Narrow webmention discovery

[snarfed]

Right now we run webmention discovery liberally on incoming activities' "targets," eg inReplyTo, mentions, etc. In practice, these are almost all fediverse/Mastodon URLs that don't accept webmentions. We should probably do one or both of:

• Load and reuse Bridgy's blocklist
• Only do webmention discovery on users' domains

[snarfed]

Simpler start here is just to cache wm discovery. Ideally we'd move Bridgy's existing caching into webutil.webmention, opt in somehow, maybe via kwarg.

[snarfed]

As for a blocklist of all existing AP instances, fediverse.observer and the-federation.info both have APIs over their node lists, easy to export all the domains. They have 46k and 41k respectively, 36k in common, 13k in just one or the other.

https://api.fediverse.observer/ :

query MyQuery {
  thefederation_node {
    host
  }
}

and https://cloud.hasura.io/public/graphiql?endpoint=https%3A%2F%2Fthe-federation.info%2Fv1%2Fgraphql

query MyQuery {
  nodes {
    domain
  }
}

[snarfed]

Added wm discovery caching. It's ok, but not great, since right now we churn instances pretty frequently. Eventually the full fediverse instance blocklist would be useful.

[snarfed]

#552 would help here.

[snarfed]

This shouldn't be happening now that we do protocol discovery.