deprecate insecure WebSockets (ws)

Nobody has claimed this yet.

I'm wondering if we should deprecate (insecure) WebSocket support:
* With the web having moved to ubiquitous TLS encryption, there's simply no websites left for which the browser would allow an unencrypted WebSocket connection. When a website is loaded via HTTPS, the browser will block all unencrypted WebSocket requests originating from that site.
* Standalone nodes could in principle still use WebSocket, as they're not bound by the browser's security model. However, there's no point in dialing a WebSocket connection to begin with: It's _much_ faster (in terms of handshake latency) to dial a libp2p TCP connection.

If I understand the issue that prompted this PR correctly, this problem wouldn't exist if we just used the SNI we use in the TLS handshake for the Host header in the WebSocket CONNECT HTTP request. We don't set the Host header currently, but that would be a much smaller fix than adding a new multiaddr component.
The values for TLS SNI and the HTTP Host Header are always identical, unless you're doing some form of [domain fronting](https://en.wikipedia.org/wiki/Domain_fronting). I'd be ok with leaving that fringe use case unsupported for now.

_Originally posted by @marten-seemann in https://github.com/libp2p/go-libp2p/issues/1829#issuecomment-1279265994_

Start with issue #1829 and review the existing WebSocket support; the payload names no files or tests. Clarify whether insecure WebSocket support should be deprecated or whether the proposed TLS SNI and HTTP Host header change is the intended scope. Done requires an agreed scope and validation plan before implementation.

Written by the indexing model from the issue text.