…ched

In this commit, I'm working on optimizing how we handle action dispatches by avoiding
repeated traversal of the `states` list. Instead, we prepare a map each time a new state
is added, allowing us to perform O(1) lookups by action type in the future. This approach
reduces complexity and improves performance.

I've tested it with benchmark.js, and here are the results:

```
class Increment {
  static readonly type = 'Increment';
}

const states = Array.from({ length: 50 }).map((_, index) => {
  @State({
    name: `counter_${index + 1}`,
    defaults: 0,
  })
  @Injectable()
  class CounterState {
    @action(Increment)
    increment(ctx: StateContext<number>) {
      ctx.setState((counter) => counter + 1);
    }
  }

  return CounterState;
});

store.dispatch() before changes x 3,435 ops/sec ±0.45% (65 runs sampled)
store.dispatch() after changes x 3,942 ops/sec ±1.21% (25 runs sampled)
```