Slicing error found by afl.rs

[StevenJiang1110]

I've used afl.rs to heavily test url2.2.0, and some slicing error may occured when specific input is provided.
To replay the panic, the test code is

let mut _local1 = url::Url::parse("arhttpsps:/.//eom/dae.com/\\\\t\\:").unwrap();
let _ = url::quirks::set_hostname(&mut _local1 ,"//eom/datcom/\\\\t\\://eom/data.cs");

I also placed the replay file at https://github.com/StevenJiang1110/afl_rust_crashes/tree/main/url/replay_url40
I hope you will check if this is an actually bug. Thanks a lot.
The simple bug report is like this.

[StevenJiang1110]

This is the totol report with RUST_BACKTRACE=full

[djc]

I've fixed the immediate issue in #655, but with URLs like this it looks like there are still some issues. For example, in the newly added tests, the before and after URLs don't have the same string representation:

thread 'fuzzed' panicked at 'assertion failed: `(left == right)`
  left: `"arhttpsps://eom/dae.com/\\\\t\\:"`,
 right: `"arhttpsps:////eom/dae.com/\\\\t\\:"`', url/tests/unit.rs:676:5

(It looks like set_hostname() ends up setting an empty hostname for the Url, which does... something weird.)

[StevenJiang1110]

I put more inputs that may cause same crash on https://github.com/StevenJiang1110/afl_rust_crashes/tree/main/url
Your fix can successfully pass all the replay files.
The input seems to have some characteristics, but I can't summarize.
I will run afl on the patched version.
Thanks a lot.

[djc]

Thank you for doing the fuzzing, it's very valuable to have someone do the work on this. Ideally we'd commit it to the repository so that we can run it again in the future after making changes, is that something you'd be able to work on?

[StevenJiang1110]

I'm now working on developing a tool to automatically generate fuzz drivers to cover all public method of a crate to find crashes. But the result is far more than ideal, for example, there are too many duplicated drivers which are a waste of processing power. I will issue an PR about all fuzz drivers until I finished this work（It doesn't take a long time, maybe several weeks）. Thanks.