Normative: change idempotency for HostImportModuleDynamically

[domenic]

This better matches HostResolveImportedModule, which only requires idempotency when it completes normally (which is analogous to the success case here).

Closes tc39/proposal-dynamic-import#80, at least on the ECMAScript side.

---

To be presented at TC39 today; I thought it'd be easier to have a concrete needs-consensus PR rather than trying to page through tc39/proposal-dynamic-import#80.

[ljharb]

Are there test262 tests for this change? (or, is it hard to test/not currently tested?)

[Snapstromegon]

@ljharb under language/expressions/dynamic-import are at least the following tests affected by this change:

• double-error-resolution
• double-error-resolution-promise

Since this PR removes required idempotency for double error, I think that those test are no longer needed.

Also double-error-resolution_FIXTURE is no longer necessary if the tests above are removed.

[ljharb]

@Snapstromegon in that case, a merged test262 PR for those changes would be needed here. Want to file it? :-D

[fatso83]

Where did this change actually end up? I had a look at the rendered output of https://tc39.es/proposal-dynamic-import/ and it only contained the old text, before this change. Then I tried searching the actual speco of ecma-262, but found hits on either pre or post text.

The most relevant sections seems to be 13.3.10 Import Calls and HostLoadImportedModule

[nicolo-ribaudo]

We unified the HostResolveImportedModule and HostImportModuleDynamically hooks, the requirement is now the second bullet point of https://tc39.es/ecma262/#sec-HostLoadImportedModule.