Add trace::ServerTiming

[yoshuawuyts]

This PR adds a first pass of the Server-Timing headers spec, introducing it as "unstable" under http_types::trace::ServerTiming. This also moves some code around in the trace submodule to accommodate more APIs being added.

Screenshot

[yoshuawuyts]

Seems to be working well!

Example code

use tide::http::trace::{ServerTiming, Metric};
use std::time::Duration;

#[async_std::main]
async fn main() -> Result<(), std::io::Error> {
    tide::log::start();
    let mut app = tide::new();
    app.middleware(tide::utils::After(|mut res: tide::Response| async move {
        let mut timings = ServerTiming::new();
        timings.push(Metric::new("mongodb".to_owned(), Some(Duration::from_millis(5)), None)?);
        timings.push(Metric::new("tide".to_owned(), Some(Duration::from_millis(10)), None)?);
        timings.push(Metric::new("db-query".to_owned(), Some(Duration::from_millis(2)), None)?);
        timings.apply(&mut res);
        Ok(res)
    }));
    app.at("/").get(|_| async { Ok("Hello, world!") });
    app.listen("127.0.0.1:8080").await?;
    Ok(())
}

Screenshot

[yoshuawuyts]

Oki, this is about the way it should be now. As this is still behind an "unstable" flag this should be good to merge. Thanks!