Fix get_dead_nodes query

[chkeita]

partial fix for #1028

[bmc-msft]

I had not considered using Timestamp (which is effectively "last updated") as a solution. Interesting.

Can this include updating the log & Error messages to differentiate between a node with an expired heartbeat and a node that never generated a heartbeat?

I'm think about this:

onefuzz/src/api-service/__app__/onefuzzlib/workers/scalesets.py

Lines 389 to 395 in 48978ce

	logging.info(
	SCALESET_LOG_PREFIX
	+ "reimaging nodes with expired heartbeats. "
	+ "scaleset_id:%s nodes:%s",
	self.scaleset_id,
	",".join(str(x.machine_id) for x in dead_nodes),
	)

and this:

onefuzz/src/api-service/__app__/onefuzzlib/workers/scalesets.py

Lines 399 to 403 in 48978ce

	errors=[
	"node reimaged due to expired heartbeat",
	f"scaleset_id:{node.scaleset_id} machine_id:{node.machine_id}",
	f"last heartbeat:{node.heartbeat}",
	],