Proxy support for WebSockets is g2g!

[thedodd]

The issue described in #95 has been thoroughly tested & is now fixed.
Shoutout to @jakule for their work on implementing a solid fix for this.

closes #53, #81, #95
closes #105 (supersedes, though adding co-authored-by to preserve attribution on original work)

Checklist

• Updated CHANGELOG.md describing pertinent changes.
• Updated README.md with pertinent info (may not always apply).

[thedodd]

@jbr thanks for your work on this. I'll update to the next release (which supports sending raw messages) once it lands. It works quite nicely!

[thedodd]

@hamza1311 this PR supersedes #105. I've given you attribution on it given that you trail-blazed much of this initial work. A few items to note:

• I've removed the backwards incompatible rename of proxy-rename to proxy-path (I'm happy to go into more details on why).
• I've preserved the work related to the nipper refactor. This was originally only an issue with Tokio, but it is actually a good refactor, and I think it helps to isolate the rest of the code from the nipper selection types.
• Everything is still on async-std/tide. I was pretty unhappy with state of affairs with Warp. The encapsulation was not great, its WebSocket support was not so hot. With this change, everything is now squarely based on (async-)tungstenite, so the backend and frontend of the proxy share the same foundation, which is quite nice and has simplified the proxy logic SIGNIFICANTLY.

The only standing drawback of this current state: no HTTP2 support. I'm not super concerned about that. Given that this is a development server/proxy, and given that H2 gracefully degrades to H1.X when needed, there should be no actual development impact. That said, when H2 support lands in Tide, we reap the benefits with no effort on our side.