Description of problem:

Neutron added bulk creation operations for certain resources like ports and subnets some time ago. This operations allow you to create N resources with a single API request.

Unfortunately, internally Neutron processes each of the N resources separately multiplying by N the amount of works that needs to be done in large stretches of the creation flow. All in all it makes it common for a 2000 port creation operation where all the requested ports are equal to take close to one hour.

This feature requests the code to be optimized so that if all the requested ports are equal, they be processed together and perform just x database transactions instead of xN transactions.

Steps to Reproduce:
1. source openrc
2. Create the following script

#!/usr/bin/env python
import sys

import os_client_config
from pprint import pprint


neutron = os_client_config.make_client('network', cloud='envvars')
nova = os_client_config.make_client('compute', cloud='envvars')

hypervisors = nova.hypervisors.list()

bulk_port_req = {'ports': []}
for hyper in hypervisors:
    if 'compute-0' in hyper.hypervisor_hostname:
        continue
    bulk_port_req['ports'].extend([{
        'binding:host_id': hyper.hypervisor_hostname,
        'project_id': sys.argv[2],
        'network_id': sys.argv[3],
        'device_owner': 'compute:kuryr',
        'admin_state_up': True,
        'name': 'kaboom',
        'security_groups': sys.argv[4].split(',')} for _ in range(int(sys.argv[1]))])

ports = neutron.create_port(bulk_port_req).get('ports')
pprint(ports)

3. chmod +x port_creator
4. Create the ports by doing:
./port_creator 2000 project_uuid network_uuid security_groups_comma_separated_uuids


Actual results:
The ports are created in x*N transactions as visible in the log.

Expected results:
The ports are created in x transactions as visible in the log.