Improve/reconsider Spec compliance of history-traversal task-source and task-queue

[gterzian]

It appears that the Spec contains two related concepts:

1. A #session-history-traversal-queue,
2. A #history-traversal-task-source

The queue under 1 appears to be a concept to be run "in parallel", and currently Servo appears to implement it via History sending messages directly to the constellation, which appears correct if only a little implicit in it's implementation. See History.traverse_history and Step 1 of#traverse-the-history-by-a-delta.

This queue doesn't seem to have an explicit task source associated with it, like the other script task queues does.

Therefore, the "history-traversal" task source isn't actually used to queue tasks on that parallel history-traversal queue, instead the task-source is used to queue task on a script event-loop, for example see the document unload steps in Step 5 of #traverse-the-history-by-a-delta.

Currently, while we do have a HistoryTraversalTaskSource, it doesn't seem like we're using it to enqueue the task mentioned at Step 5, and also the task source doesn't actually implement TaskSource and appears un-used.

We could consider implementing TaskSource for it, and use it in Step 5 of #traverse-the-history-by-a-delta

cc @cbrewster @AgustinCB

[gterzian]

Related whatwg/html#4054

[gterzian]

Closing in favor of #10994