proc: check that PID still exists on signal operations

slack context: https://iterativeai.slack.com/archives/C03JS2V4MQU/p1693318074581219?thread_ts=1693317855.876959&cid=C03JS2V4MQU

When `os.kill` returns an OSError we should verify that the PID we are trying to signal is still valid, and just update the task as failed/finished if the PID is no longer valid