reduce default timeouts
Right now the timeouts for a teuthology job are quite long. Historically this was because the timeout period included the attempt to lock nodes, which could take a while.
Now, though, we have a "waiting" state!
So instead of an extremely large timeout (Zack said he thought it was 36 hours) we can have a shorter one (12 hours, maybe?) that doesn't begin until we exit the waiting state.
That would help with situations like #11314 where a broken task (that fails and hangs without noticing) will lock up all the machines in a job until the full teuthology timeout has passed.