sched: qhmp: optimize power_delta_exceeded() for !sched_enable_power_aware

The relative power cost difference between higher capacity and lower
capacity clusters can be less than the default value of
sysctl_sched_powerband_limit_pct i.e 20. As there is no sysctl knob
to modify this threshold, we end up ignoring the power cost difference
between the clusters. This results in power regression.

When power aware scheduling (sched_enable_power_aware) is disabled,
all the CPUs in a cluster will have the same power cost. Hence
we can optimize power_delta_exceeded() to return true if there is
any difference in the power cost. This optimization also fixes the
above problem.

Change-Id: Id21495f23e439386edbefd13d0c4ad1d49b02a5e
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>