feat(workerd): remote eval api

[hi-ogawa]

Inspired by pupetter/playwright's eval https://playwright.dev/docs/evaluating, this would be probably the simplest but somewhat general enough remote runner execution API.

export type WorkerdDevEnvironment = 
  DevEnvironment & {
    api: {
      dispatchFetch: (entry: string, request: Request) => Promise<Response>,
      // [ NEW ]
      eval: (entry: string, fn: (entryModule: unknown, ...args: Args) => T, ...args: Args) => Promise<T>,
    }
  }

If we customize (de)serialization (e.g. by devalue, turbo-stream), this API might be powerful enough for common use cases.

TODO

• demo
  • workerd repl?
• (de)serialization customization
  • serialization library can be even imported by runner, so it doesn't need to be bundled?
  • by specializing Request and Response cases, dispatchFetch can be replaced by a general eval API?