[Enhancement] Don't override stdlib for force_resolve implementation

[syeopite]

Is your enhancement request related to a problem? Please describe.

The current force_resolve implementation relies on patching the HTTP::Client in stdlib. It really isn't good practice.

Describe the solution you'd like

I've recently been made aware that it is possible to force the HTTP::Client to resolve a specific IP family without overriding the stdlib implementation.

See crystal-lang/crystal#14196 (comment)

The downside is that we'll lose the automatic reconnect feature.

Describe alternatives you've considered

Create a new http client that inherits from stdlib with force_resolve implementation

Additional context

[SamantazFox]

Thanks for opening an upstream issue for that!

[syeopite]

Alternatively, we could just create a new client object that inherits from the stdlib HTTP::Client