Only take reference to a looked up rule when needed.
This reduces the total CPU utilization of rule_ref/unref calls by 80%,
from 5% of total server CPU capacity to 1% in a netperf TCP_CRR
test stressing the userspace.
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com> Acked-by: Ethan Jackson <ethan@nicira.com>