enhance `libvirt ssh` timeouts and feedback

[cgwalters]

Right now I think our ssh timeouts are a bit slow I think, we should speed those up and also ensure feedback.

Yeah ideally the vsock would always be there, but there's environments and OSes it doesn't work with. One tangential thing I'd like to do here though is to always ensure we get more feedback out of the libvirt case. Maybe worst case we can scrape the console if we can rely on the OS writing to hvc0.

[cgwalters]

Prior art here is how cosa run from coreos-assembler scrapes the console by default.

I think it'd be interesting to have say --journal-log that scrapes the journal too via an injected systemd unit, which coreos-assembler also has code to do. An interesting question is where that log goes and if it's infinite or more ringbuffer format. I would lean towards the latter.

[gursewak1997]

What I have worked so far locally: enhancing the bcvk libvirt ssh command with faster timeouts and better feedback.

• The default SSH timeout was reduced from 30 seconds to 5 seconds for quicker failure detection.
• When SSH isn't ready yet, the command now implements a retry mechanism (3 attempts) that scrapes the VM's console output via virsh console --force between retries, showing boot progress like systemd targets and service starts. This provides users with visibility into what the VM is doing instead of just hanging silently.
• Also fixed critical resource management issues to prevent zombie processes and blocking I/O.

Running some tests now, I see them confirming fast connections when ready, proper console feedback during retries, and clean error handling so far.

EDIT: Looking into the cosa run context