Add support to specify pre-resolved IP addresses and avoid additional DNS lookup

[rjobanp]

Necessary to support this PR MaterializeInc/materialize#26186 by allowing us to specify multiple pre-resolved IP addresses to avoid a DNS lookup inside the library.

I wanted to know what you thought of the API changes before opening the upstream PR. Since we recently also added support for an explicit TLS hostname override https://docs.rs/mysql_async/latest/mysql_async/struct.SslOpts.html#method.with_danger_tls_hostname_override it seems a bit clunky to keep the hostname field on the new HostPortOrUrl::ResolvedHost enum type I introduced, but without that field it's unclear what we should return from get_ip_or_hostname, which is behind a public method returning a &str

[benesch]

I wanted to know what you thought of the API changes before opening the upstream PR.

I feel like we should be able to do a big unification/simplification? It seems like we could remove with_danger_tls_hostname_override, in favor of the new resolved_ips API?

[rjobanp]

I feel like we should be able to do a big unification/simplification? It seems like we could remove with_danger_tls_hostname_override, in favor of the new resolved_ips API?

I'm not sure if the ergonomics would be the same. If someone still wanted this library to resolve DNS and to override the TLS hostname, there wouldn't be a way to do that with the resolved_ips API. So I think we still need both, but I will incorporate your other comments to avoid introducing the 3rd enum variant of HostPortOrUrl

[benesch]

If someone still wanted this library to resolve DNS and to override the TLS hostname, there wouldn't be a way to do that with the resolved_ips API.

Yes, agreed, it would somewhat limit the ergonomics. Arguably fine since before last month no one had wanted the with_danger_tls_hostname_override API, and as of this month we don't need it either. I don't feel strongly though, and happy to defer to you/upstream on whatever is most palatable.

[rjobanp]

Sounds good. I'll defer to upstream when I open a PR for this branch there and ask if it would make sense to remove the with_danger_tls_hostname_override API.

[benesch]

This looks great to me!