kernel/mutex: Clean up k_mutex_unlock()

Recent changes to the scheduler API means we can simplify this
further: move the assignment to mutex->owner outside the if(), which
removes the need to have an else clause (which just set that field to
NULL when the new_owner was already NULL); and we can likewise move
the irq_unlock() outside the block.

Signed-off-by: Andy Ross <andrew.j.ross@intel.com>