Support negative time deltas in BWE

[alexlapa]

So first of all we should be able to handle negative send deltas in ArrivalGroup as discussed before: 0, 1, 2

So this is pretty much what this PR does so far, but it might be extended as discussed here.

So, I can leave it as it is, right now it only affects packet/bwe and these new types are only used where it's expected that negative time deltas are possible. Alternatively:

1. Always prefer using these new types instead of std's in BWE? They work pretty much like libwebrtc's Timestamp and TimeDelta so they should fit.
2. Also use it to cover similar concepts as not_happening(), already_happened()? I haven't really though about it but first thing that comes to mind is just using SuperInstant instead of those Instants that can be already_happened or not_happening with some simple conversion SuperInstant::as_instant() -> Instant that will just return already_happened || Exact(Instatnt) || not_happening. Might also just put a Deref -> &Instant on it.

So let me know what you think about this.

Also I've returned that ArrivalGroupAccumulator test on libwebrtc captured data, but now it also runs on negative send deltas. Ill remove it before merge.

[algesten]

Looks good.

Yeah. So I want to replace as much as possible of the time_hacks.rs. Ideally I'd like to call these things Instant and Duration (instead of SuperInstant and TimeDelta). The names deliberately shadow the std-lib ones because they should function 99% like the std-lib does. That means a casual reader of the code should not be surprised by new time types.

Our external API is however std-lib Instant and Duration, which means there's work to go through exactly all the internal code where this would affect.

I'd prefer if the SuperInstant/TimeDelta could be moved to a separate PR that handles that and leave this PR to only deal with the BWE consequences.

[thomaseizinger]

Our external API is however std-lib Instant and Duration, which means there's work to go through exactly all the internal code where this would affect.

Might be worth adopting something like https://github.com/obi1kenobi/cargo-semver-checks to ensure changes to the public API are automatically linted!

[alexlapa]

Might be worth adopting something like https://github.com/obi1kenobi/cargo-semver-checks to ensure changes to the public API are automatically linted!

Well in this specific case E0446 should be enough I suppose.

[algesten]

Given the findings in #617. Let's go further with this PR by making the negative-time handling variants local to the BWE module. SuperInstant is maybe a questionable name, but TimeDelta is good. Let's isolate them to bwe module so they don't spread across the rest of the code base for now.

And again, sorry for making you doing a wild goose chase in a direction that doesn't work.

[alexlapa]

SuperInstant is maybe a questionable name

That's how a similar thing was named before. I would probably go with Timestamp since thats how it's named in libwebrtc. Let me know if you'd like to have a different name.

[algesten]

SuperInstant is maybe a questionable name

That's how a similar thing was named before. I would probably go with Timestamp since thats how it's named in libwebrtc. Let me know if you'd like to have a different name.

Let's rename and stick closer to libWebRTC.

[algesten]

LGTM! Some minor comments

[algesten]

@alexlapa Thanks!!!