fix(ffi): hold the room weakly from its own RPC method handlers

[LautaroPetaccio]

The problem

FfiParticipant::register_rpc_method cloned Arc<RoomInner> into the handler it registers on the SDK:

let room: Arc<RoomInner> = self.room.clone();
local.register_rpc_method(method.clone(), move |data| { ... room ... });

That handler is stored on the room's own RPC server. The ownership chain is Arc<RoomInner> → Room → Arc<RoomSession> → rpc_server.handlers (livekit-rpc/src/server.rs:50) → the closure → back to Arc<RoomInner>.

Nothing unregisters the method during teardown:

• RoomSession::close fails pending client RPCs but never touches rpc_server.handlers.
• FfiRoom::close does not touch it either.
• FfiServer::dispose only clears the handle map, which drops one Arc out of a self-sustaining cycle and releases nothing.

So a single RegisterRpcMethodRequest makes the room outlive dispose(), keeping the engine, its peer connections and the PeerConnectionFactory resident for the rest of the process — precisely what the dispose() doc comment says it exists to prevent. Repeated connect/dispose cycles leak a full room each time. Only an explicit UnregisterRpcMethod per method broke the cycle.

The fix

Capture a Weak and upgrade per invocation, as AGENTS.md requires for a callback stored by the object it captures:

A callback stored by an object must not capture a strong clone of that same object.

An invocation arriving once the room is gone now fails cleanly with an application error rather than resurrecting it.

Verification

Adds __lk-e2e-test to livekit-ffi, mirroring the feature in livekit, and a test under it that drives the real FFI path: connect a room, complete the room-event ready handshake, register a method, dispose(), then assert the room's internals are released.

• Before: fails — "the registered RPC handler retained the room after dispose".
• After: passes.

Since making the capture weak is the only change, that strong capture was by definition the retaining reference.

All 17 livekit-ffi tests pass, including the existing dispose_cleans_up_resources. The default build (feature off) compiles, and cargo check --features __lk-e2e-test --tests at the workspace root — how CI invokes the suite — resolves correctly with the new feature.