[cellv2] performance issue when querying compute nodes in a cell

[xhzhf]

Description
===========
we used rally to create instances concurrently.
we found it is slow and Nova queryed compute nodes in cell0

Steps to reproduce
==================
use rally to create instances concurrently.

Expected result
===============
Nova should not query compute nodes in Cell0

Actual result
=============
Nova query compute nodes in Cell0

Environment
===========
1. Exact version of OpenStack you are running. See the following
  list for all releases: http://docs.openstack.org/releases/
   openstack-nova-scheduler-16.0.3-1.el7.noarch
   openstack-nova-placement-api16.0.3-1.el7.noarch

2. Which hypervisor did you use?
   KVM. We have 2000 hosts

2. Which storage type did you use?
   Our own storage device

3. Which networking type did you use?
   Our own network device

Logs & Analysis
==============
4916532:2017-12-14 13:59:45.473 1581 DEBUG nova.scheduler.host_manager [req-2d5fb838-d92c-49b9-a32d-0791185152c4 04433cd94b5f4982b224ff0f2afbf3f3 16cab611766742398402a19b9a58d71f - default default] Getting compute nodes and services for cell 00000000-0000-0000-0000-000000000000(cell0) _get_computes_for_cells /usr/lib/python2.7/site-packages/nova/scheduler/host_manager.py:608

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/529504

[OpenStack Infra (hudson-openstack)]

Change abandoned by Balazs Gibizer (<email address hidden>) on branch: master
Review: https://review.opendev.org/529504
Reason: This is a pretty old patch with negative review. Feel free to restore it (or ask gibi on irc to restore it) if you still working on it.