hub: upstream improvements for identity

[shikokuchuo]

Upstream Improvements

1. Expose set_actor in automerge WASM bindings (automerge/automerge)

The Rust automerge crate has doc.set_actor(ActorId) — a single field assignment with no allocation or copying. But the WASM bindings (automerge_wasm_bg.js) don't expose it. The only way to change the actor from JS is Automerge.clone(doc, { actor }), which copies the entire document.

Adding setActor(actor: string) to the WASM Automerge class is a one-line binding to the existing Rust method. The JS layer would then expose it as Automerge.setActorId(doc, actorId) or similar. This would let us replace:

// Current: clone entire document to change one field
handle.update(doc => Automerge.clone(doc, { actor: actorId }))

// With set_actor exposed: mutate in place, no allocation
handle.update(doc => { Automerge.setActorId(doc, actorId); return doc })

Priority: low. The clone is safe (no sync side-effects, old doc is GC'd) and cheap for our document sizes. But it's the right fix long-term.

2. Add actor option to Repo.create() / Repo.find() (automerge/automerge-repo)

Ideally, Repo.create() and Repo.find() would natively accept an actor option:

repo.create<T>({ actor: hexActorId })
repo.find<T>(docId, { actor: hexActorId })

This would eliminate the update+clone workaround and all its downsides:

• Noise in history: create() writes one change with a random actor before the clone switches to the real one. That random actor is permanently baked into the document.
• Performance: Automerge.clone() copies the entire document just to change the actor ID.

Migration would be straightforward — update the createDoc/findDoc helpers to pass the native actor option and remove applyActorId.