Refactor Middleware to Logger

[lexnv]

Change the name of the Middleware traits (HTTP / WS) to better reflect
its purposes.

Closes #843.

[jsdw]

I'm not certain on the word Metrics on it's own but I could live with it! Logger was another word that came to mind :)

[niklasad1]

yeah, I'm not a fan of using HttpMetrics either we should really have like HttpRpcLogging/Metrics because we need to make it clear that you can't really do any logging/metrics on the underlying HTTP/WS instead it gives timings on the specific JSON-RPC calls except some edge-cases just as on_connect but perhaps that can be removed with tower middleware...

[jsdw]

I mean, since on_connect and on_disconnect exist, we do allow some logging based on the whole http/ws connection too?

My fave really would be HttpLogger or WsLogger; the user would then impl eg HttpLogger for some struct and it would do some sort of logging of http things (mainly the rpc request bit but also the whole http cycle via the on_connect/on_disconnect too right?).

impl HttpMetrics for SomeStruct feels a bit wonky/less clear to me (and, like, we might log metrics or we might log some other stuff, but logging something is really the thing that this trait enables, basically).

[niklasad1]

with that said, shouldn't we expose the entire HTTP request then instead of just the headers?

then I'm on board with HttpLogger

[jsdw]

If it's easy to do then I don't see why not!

[niklasad1]

LGTM, nice work