Sync interpreter

The synchronous interpreter lives in xstate-core (re-exported by the facade). It drives the sans-I/O step engine inline: no runtime, no tasks, no I/O. This is the interpreter for request / response services, tests and the CLI.

#![allow(unused)]
fn main() {
use xstate_statemachine::prelude::*;

let machine = Machine::from_json(r#"
  {"id": "door", "initial": "closed",
   "states": {"closed": {"on": {"OPEN": "opened"}}, "opened": {"on": {"CLOSE": "closed"}}}}
"#)?;
let mut door = SyncInterpreter::new(machine, MachineLogic::dynamic());
door.start()?;
door.send("OPEN")?;
assert_eq!(door.state_ids(), ["door.opened"]);
Ok::<(), xstate_statemachine::core::Error>(())
}

The API

• SyncInterpreter::new(machine, logic) uses a RealClock; with_clock(machine, logic, clock) takes a SimulatedClock so tests can move time with advance(&clock, ms) instead of sleeping.
• start() runs the initial macrostep and reports missing implementations. send(event) takes a &str, an Event with a payload, or a typed event, and returns a Receipt (changed, unhandled, the actions that ran, status() and state_ids()).
• state_ids() lists the active leaf state ids (<machine id>.<path>); context() returns the current context; status() is running, done, error or stopped.
• snapshot() / snapshot_json() capture the instance; SyncInterpreter::from_snapshot(..) restores one (see snapshots).
• observe(Arc<dyn Observer>) attaches observers (metrics, inspector, transition log).

The core crate also exposes the bare step function and the pure API (initial_transition, transition) for programs that want the algorithm without an interpreter, and graph utilities (stub_logic, reachable_states, shortest_paths, simple_paths) for analysis. The async equivalent is spawn on tokio; the two share one algorithm (ADR-0004).