Queue tasks to resolve or reject promises

[twiss]

Addresses #368 by creating a "crypto task source" and queueing tasks on that task source in order to resolve or reject all promises created in response to calls to methods of SubtleCrypto while running in parallel, to prevent race conditions.

In addition, create or convert to ECMAScript objects in those tasks, inside the top-level methods, rather than inside the internal operations; and let all internal operations consistently return IDL objects instead of ECMAScript objects (deriveBits already did this, but the others did not).

Finally, clarify some of the affected spec language and the types involved.

@dontcallmedom and/or @tidoust, could you please additionally review the last commit, in particular?

[dontcallmedom]

I reviewed only the last commit which looks correct to me, but I would trust @tidoust 's review more than mine

[annevk]

This generally looks good to me, but it would be nice to at some point modernize more of the writing style.

[annevk]

This is a rather confusing way of writing algorithms.

[twiss]

Yeah, I agree. However, I also think it's a somewhat useful shorthand here because to get rid of this entirely, we would need to pass the promise and the realm to the internal operations, which would then each need to queue a task whenever it encounters an error or reaches a result.

Conceptually speaking, it's somewhat similar to saying

return Promise.try(async function() {
  // remaining steps of this algorithm
});

But perhaps there's a better way of saying that?

[annevk]

There is https://webidl.spec.whatwg.org/#an-exception-was-thrown but it's kinda weird to throw exceptions while in parallel as they are ECMAScript objects.

[annevk]

Infra keeps the quotation marks outside of the code markup. "{{KeyFormat/raw}}" works for that.

[twiss]

OK, yeah. I'll make a separate PR for that as this is used all over the place.

[annevk]

Seems the spec is in contradiction to the last paragraph of https://w3c.github.io/webcrypto/#conformance

[twiss]

I think that text was more relevant before #309, in which I put many quotes outside <code>.

... I was gonna say ".. but apparently not all of them", but it seems the markup works fine as is, as well: https://w3c.github.io/webcrypto/#SubtleCrypto-method-importKey (which has similar markup as this one).