Add tests for interaction between import() and microtask queue

[nicolo-ribaudo]

The HTML algorithms for dynamic import (starting from HostImportModuleDynamically) only schedule a task a when actually fetching the corresponding files, and it never happens if the imported module (and all its dependencies) have already been loaded (when step 5 of fetch a single module script doesn't happen and step 6 returns early).

I didn't know how to test "it doesn't schedule a task", so the tests are written using a loop that schedules many microtasks and checking if all those microtasks are run or not before that the import() promise is resolved.

---

Browsers implement this inconsistently. This PR adds three new tests, and some of them fail:

Test	Firefox & Chrome	Safari
import() should not drain the microtask queue if it fails during specifier resolution	:heavy_check_mark:	:heavy_check_mark:
import() should not drain the microtask queue when loading an already loaded module	:x:	:heavy_check_mark:
import() should drain the microtask queue when fetching a new module	:heavy_check_mark:	:heavy_check_mark:

It's possible that the Firefox/Chrome behavior is because the HTML spec uses a rare "asynchronous completions" wording which isn't defined anywhere. However, "asynchronous completions" are used for all the test cases, so I would expect the 1st to always behave like the 2nd (since they both only involve asynchronous completions and no fetching). I also opened https://github.com/whatwg/html/pull/8264 to move away from that wording.