Author-assigned labels to reference labeller service definitions

[Bossett]

Is your feature request related to a problem? Please describe.

Related to #3870 and #3871 - there is conflict between labels from different services and author labels.

Describe the solution you'd like

Author self-labels should contain a reference to a labeller (like the official moderation service), and the user should be able to pick from labels from services that they use (maybe a subset of - could be defined in the labeller service definition 'canSelfLabel' or something).

This will remove the possibility of meaning drift between how an author label is described, and a labeller (i.e. for things like 'nudity'), and allow users and labellers to interact in interesting ways (i.e. a user can say "I'm going to label all my spiders from the spider labeller, so go sub to that if you don't want to see them").

These structures already exist in the protocol - you could just set a src for an author label.

This also allows a labeller service to 'undo' an author label by pushing a negate label, because the conflict resolution is built in (i.e. author trusts labeller, label contains labeller did). If you want to get really clever, this could then be handled by ozone to allow self-unlabelling by confirming an appeal came from the author.

Describe alternatives you've considered

See #3870 & #3871

[Nayelianne]

I'd love this, specially because it'd allow users to ease the burden of certain labelers (like the phobia and spoiler ones). Having no delay to apply the moderation label (since the author themselves would be applying it), would also mean more protection for users who are subscribed and serve as a good workaround until we have more customizable content warnings.