In case Taker.initialize() fails for any reason such as no
liquidity, stallMonitor() will not wake up for 20*maker_timeout_sec;
so, in this change, IF the client schedule has length one, we
immediately figure the on_finished_callback with failure parameters
so that the reactor is stopped without waiting.
This change does not apply for the general multi-step schedule, as
that is principally for tumbler-style algos which are intended to
aggressively keep trying in case of any failure.