Issue with boot_vm_attach_detach_volume at scale

When running boot_vm_attach_detach_volume with times 1000 and concurrency 16, 2 iterations had the following error.

neutronclient.common.exceptions.BadRequest: Bad floatingip request: Port *** does not have fixed ip ***. Neutron server returns request_ids:[***].

I think this might be an issue in the Rally code for _boot_server_with_fip, as I have seen this issue once before in another scenario which boots servers with floating IPs.