sched/fair: Refactor cpu_util_without()

Except the 'task has no contribution or is new' condition at the
beginning of cpu_util_without(), which it shares with the load and
runnable counterpart functions, a cpu_util_next(..., dst_cpu = -1)
call can replace the rest of it.

The UTIL_EST specific check that task util_est has to be subtracted
from the CPU one in case of an enqueued (or current (to cater for the
wakeup - lb race)) task has to be moved to cpu_util_next().
This was initially introduced by commit c469933e
("sched/fair: Fix cpu_util_wake() for 'execl' type workloads").
UnixBench's `execl` throughput tests were run on the dual socket 40
CPUs Intel E5-2690 v2 to make sure it doesn't regress again.

Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>