No implementation calls #resume-applying-the-traverse-history-step entirely async

[farre]

What is the issue with the HTML Standard?

In the case where a history traversal navigation is intercepted, #resume-applying-the-traverse-history-step is expected to be called asynchronously, but no implementation does this entirely. WebKit performs synchronous steps, as well as Chromium, and Gecko is in the process of adding synchronous steps.

Tests concerned are, amongst others. navigation-api/navigate-event/intercept-same-document-history-back.html, but also tests listed in Bug 1992939. This bug reflects that for Firefox this broke Reddit, which prompted us to do this spec breaking change.

The root cause was that the current navigation entry needs to be updated before event handlers run.

[farre]

@zcorpan, @avandolder, @noamr

[noamr]

cc @natechapin

[noamr]

@domfarolino

[farre]

It turns out that this is worse thank I thought. The async call to apply the history step falls down into the wait for all making several scroll tests not spec correct as well. Basically all tests doing

await navigation.back().finished
assert_true(some scrolling condition) 

are wrong, and needs an await new Promise(resolve => setTimeout(resolve, 0)) before the assert to have a chance to be spec compliant, even though that is probably not enough.