Reorganize types and traits to deprioritize Vec streams

[frankmcsherry]

Timely makes it easy, arguably too easy, to default to Vec container streams. They are a fine default, but the software architecture should have them as an opinion layered atop a common core, rather than the core routing around their opinions.

This PR attempts to deprioritize the vector-backed streams, renaming them StreamVec and migrating their extension traits to a vec module under operators, a peer to core rather than the default entry point. This is fairly disruptive, though there are pretty easy fixes in each case. Usually one can get what one wants by importing the vec flavored traits, though also using the core traits is not so bad with the .container::<Vec<_>>() annotation. Probably not how you want to start, but introducing the trade-off seems helpful.

The documentation needs revision, and many comments are no longer accurate. Many references to Stream still exist, StreamCore is still a thing, and tbd on just how terrible it looks to use. Much of the annotation annoyance comes from demo code that doesn't do anything, and the examples where something of consequence happens largely didn't need much change (the container types are clear, because nothing is so general that it works for all/many containers).

[antiguru]

Looks good! Mostly a cursory look, but nothing fundamentally wrong here.