Shut down DAP server gracefully on disconnection

[lionel-]

Branched on #1016.

Follow up to posit-dev/positron#11407 and posit-dev/positron#11497 where we disabled the exthost restart test on Windows (Slack discussion at https://positpbc.slack.com/archives/C06DNFJSHPD/p1769113239000079).

Not sure what this only surfaced on Windows, but the cause was an Ark crash due to sloppy Unwraps on connection events. This PR makes sure connection errors are propagated, logged, and cause the DAP server to shut down gracefully.

It was a bit tricky to debug because the logs were not showing the crash. The only clue that surfaced was a message in the Console:

and these messages in the Supervisor logs:

10:00:05 [DEBUG] (2) kcserver::kernel_session::process: End of output stream (kind: Stderr)
10:00:05 [DEBUG] (18) kcserver::kernel_session::process: End of output stream (kind: Stdout)
10:00:05 [INFO] Child process for session r-68064bdf exited with status: exit code: 0xc0000409
10:00:05 [DEBUG] (17) kcserver::kernel_state: [session r-68064bdf] status 'idle' => 'exited' (child process exited)
10:00:05 [DEBUG] (17) kcserver::heartbeat: [session r-68064bdf] Stopping heartbeat monitor (exit event signaled).

0xc0000409 is the code for STATUS_STACK_BUFFER_OVERRUN.

Attaching Ark with a debugger then restarting the exthost gave enough time to the process to emit log messages before exit, revealing the DAP panic.

[DavisVaughan]

As usual, I'm a very big fan of less unwrap()s 😄

[DavisVaughan]

So is it getting dropped on the sender side when the ext host restarts suddenly?

[lionel-]

I don't think we've hit that code path, it's a defensive fix. It would happen if the DAP comm closed before the DAP connection I think.