Multiple new stream adapter updates

[jasnell]

A number of cleanups and more improvements.

For the write adapter, when writing to an internal stream (think fetch, identity transform stream, etc)... we will directly support writing ArrayBuffer, ArrayBufferView, SharedArrayBuffer like we do currently, but we also support writing strings directly (as utf8 bytes) which will avoid hving to create a TextEncoder. And we'll also support writing sync iterables, e.g.

const { writable, readble } = new IdentityTransformStream();

const writer = writable.getWriter();

await writer.write('hello');  // will work

await writer.write(['hello', 'there']);  // will work

function* gen() {
  yield 'hello';
  yield 'world';
}
await writer.write(gen());  // will work

Why you ask? Pretty simple.. writing multiple chunks at once will help avoid the more expensive KJ/JS promise loop.

[harrishancock]

Suggested change

	// The returned promise will never resolve with more than maxBytes
	// The returned promise will never resolve with more than maxBytes.

[harrishancock]

Here and elsewhere, you could use JSG_TRY / JSG_CATCH for new code now, if you want.

[harrishancock]

Do we need to worry about old workers still? My understanding it's a moot point whether we detach or not, because we have to use a non-V8 heap buffer anyway, but if detaching versus not detaching is a script-visible side effect, could removing this break anyone?

[harrishancock]

Nit: would be good to note the current limit in the error message

[harrishancock]

Nit: could be JSG_TRY