Some L3 RPCs are time-consuming especially get_routers

[LIU Yulong]

Examples:
http://logs.openstack.org/11/669111/4/check/neutron-tempest-plugin-dvr-multinode-scenario/dc3af26/controller/logs/screen-q-l3.txt.gz#_Jul_07_04_18_11_791730
http://logs.openstack.org/11/669111/4/check/neutron-tempest-plugin-dvr-multinode-scenario/dc3af26/controller/logs/screen-q-l3.txt.gz#_Jul_07_04_18_45_507205
http://logs.openstack.org/11/669111/4/check/neutron-tempest-plugin-dvr-multinode-scenario/dc3af26/controller/logs/screen-q-l3.txt.gz#_Jul_07_04_20_04_023626
http://logs.openstack.org/11/669111/4/check/neutron-tempest-plugin-dvr-multinode-scenario/dc3af26/controller/logs/screen-q-l3.txt.gz#_Jul_07_04_20_34_698863

Some possible reasons:
1. neutron server is busy while it can not process this RPC
2. DB slow query
3. L3 agent thread (context) switching

[Bernard Cafarelli (bcafarel)]

Maybe a bug to bring to the performance meeting?
http://logs.openstack.org/11/669111/4/check/neutron-rally-task/e48e2e7/results/report.html.gz

If some of these calls are large offenders time-wise, we could check the osprofiler results like http://logs.openstack.org/11/669111/4/check/neutron-rally-task/e48e2e7/results/report.html.gz#/NeutronNetworks.create_and_list_routers

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/669111
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=9c4bd4bd9a718d17973a9962397243c5a3d107b1
Submitter: Zuul
Branch: master

commit 9c4bd4bd9a718d17973a9962397243c5a3d107b1
Author: LIU Yulong <email address hidden>
Date: Thu Jul 4 16:40:10 2019 +0800

    Add a common timecost wrapper

    And set it to all the L3 RPC functions. Move to neutron-lib
    if it will be widely used corss other projects.

    Related-Bug: #1835663
    Change-Id: Ie7743db097fd45df432af341470336d6a5662c6f