Tighten up multiArgs TypeScript type constraints

[AaronFriel]

This PR modifies the type constraints by using rest elements/spread types by default, and treating the single-argument (non-multiArgs) case as special by passing it to Emitter as [EmittedType] (note the brackets, making it a single element tuple).

This provides better ergonomics on iterator, e.g.: this case with Express:

      const testRouteIterator = pEvent.iterator<[Request, Response]>(app as any, 'test', {
        multiArgs: true,
      });

      for await (const callback of testRouteIterator) { // callback: [Request, Response]
        const [req, res]: [Request, Response] = callback;
      }

Comparison

Relative to what version 3.0 emits for types:

Using iterator<[Request, Response]>

      const testRouteIterator = pEvent.iterator<[Request, Response]>(app as any, 'test', {
        multiArgs: true,
      });

      for await (const callback of testRouteIterator) { // callback: [Request, Response][]
        const [req, res]: [Request, Response] = callback;
      }

This doesn't make sense, because it's not an array of [Request, Response] tuples.

Using iterator<Request, [Response]>

      const testRouteIterator = pEvent.iterator<Request, [Response]>(app as any, 'test', {
        multiArgs: true,
      });

      for await (const callback of testRouteIterator) { // callback: (Request | [Response])[]
        const [req, res]: [Request, Response] = callback;
      }

This doesn't make sense because the first argument is never a Response, and the second argument is never a Request, but indeed, this type checks, but it cannot possibly be true:

      for await (const callback of testRouteIterator) { // callback: (Request | [Response])[]
        const arg0: Request | [Response] = callback[0];
        const arg1: Request | [Response] = callback[1];
        const arg2: Request | [Response] = callback[2];
        const arg3: Request | [Response] = callback[3];
        const arg4: Request | [Response] = callback[4];
      }

[sindresorhus]

// @BendingBender

[BendingBender]

I think this is a breaking change...

[AaronFriel]

@BendingBender Correct, I don't usually bump version numbers as part of PRs but if I had, this would suggest a bump to 4.0.0.

[AaronFriel]

@sindresorhus @BendingBender I just want to say thank you for your speedy reviews and approval, I know it is too often a thankless job. :)