[vsock] wait for device readiness

[papertigers]

Fixes #1110

[iximeow]

nice fix, this is an unfortunately subtle interaction and unfortunately hard to test...

[iximeow]

remarking for onlookers: we'd talked about setting this up for a Tokio watch or something where we can discover that the sender went away, or that the sender actually wanted to tear down this loop without ever starting it, but @papertigers wants to refactor some of this now that we've seen how this ends up actually used anyway.

so the structure of the event loop here even might not survive too long, and the interim of accumulating an idle thread when propolis-server is used in ways we don't use it seems fine for now.

[papertigers]

Note that while testing this with @iximeow and @leftwo we ran into #1115

We were able to confirm this fix by duplicating line 607 at line 610

propolis/bin/propolis-server/src/lib/vm/state_driver.rs

Lines 596 to 610 in fcf37ae

	// Send synchronous start commands to all devices.
	for (name, dev) in objects.device_map() {
	info!(self.log, "sending start request to {}", name);
	let res = dev.start();
	if let Err(e) = res {
	error!(
	self.log, "device start() returned an error";
	"device" => %name,
	"error" => %e
	);
	return VmStartOutcome::Failed;
	}
	}

Before the change in 4a309f3 we crashed, and after 4a309f3 was applied we no longer crashed.