rmove DNS out of rama-tcp

[GlenDC]

Instead of handling DNS resolution implicitly, we should provide a dedicated connector service backed by a DNS provider.

Together with the work in #840, this would make connectors such as TcpConnector much simpler and more composable.

The idea is that connectors are stacked, each with a single responsibility:

... other connectors
|
ProxyTargetConnector
  Uses the proxy target if present as the connection target
|
DnsAddressResolverConnector
  Resolves the target if it is a domain name
|
TcpConnector
  Only accepts a socket address (or something convertible to it).
  If the target is still a domain name, it should fail.

This keeps each connector simple and explicit.

Benefits

• Clear separation of responsibilities between connectors
• Simpler connectors such as TcpConnector
• More flexible connector stacks
• Enables deeper control over the DNS stack in the future (for example allowing rama-dns to rely on rama-tcp)

Implications

For higher level utilities such as EasyHttpWebClient, the stack builder can inject the required connectors automatically depending on the needs.

This approach also avoids leaking connection related data through extensions.

Extensions

We should stop inserting ProxyAddress into extensions.

Instead we should inject:

• ProxyTarget
• ProxyCredentials

ProxyAddress can still exist as a type, but it would no longer have a Rama defined meaning inside extensions and would not be injected automatically. Users can still insert it themselves if they want.

[GlenDC]

In terms of current priorities, I plan to tackle the following issues in roughly this order:

• Finish Apple NE support and integrate it for one partner, ensuring a full transparent proxy cycle works (TCP only for now. UDP is not needed yet since we do not support QUIC or HTTP/3). (#836)
• Proper multipart support in Rama (server and client). (#139, #647)
• Windows WFP support in Rama (kernel driver and related components)
• URI rework. (#724)
• Revisit graceful shutdown. (#830)
• This issue together with #840

As you can see, there is still quite a bit of work planned. If you decide to take on these two issues, it would only make sense if you can move relatively quickly and are motivated to take them from start to finish. Otherwise it is probably better if I handle them myself.