adapt to vfio-pci driver

add judge before use slices's index

fix IP/route transfer on node reboot

add master check when a node adding to a cluster and config sb/nb add…

update installation scripts

do not delete statefulset pod when update pod

Merge changes from install.sh into install-pre-1.16.sh.

do not disable offload when the hw_offload parameter is true

fix: node route should filter out 'vpc'

docs: show openstack docs and docker image status

feat: lb switch

fix: clean up gateway chassis list for external gw

some fixes in vlan initialization

add doc for openstack/kubernetes hybrid deploy

when external gateway's type is changed from
'distributed' to 'centralized' or external-gw-nodes
list changes, it needs to clean up the existing
chassis gateways.

fix: check crds when controller start

[WIP] vpc load balancer

Fix multicast when nic type is OVS internal port

1. do not change MAC address of provider interface;
2. add routes to OVS bridge in order by scope;
3. simplify error checks.

fix: lsp may lost when server pressure is high

When server under high pressure, etcd and ovn-db changes leader frequently, sometimes the pod has the allocation ready annotation but the lsp is lost.

Some possible reasons are:
1. When ovn-db changes the leader, the lsp data is not propagate to all instances and leader change, the lsp is not persistent but we have already write back the annotation.

2. The lsp has been written to all ovn-db instances, but etcd changes the leader the apiserver takes long time to process the annotation write request. During this period the gc process is triggered and recycles the lsp before the annotation write success.

To resolve it:
1. Add --wait=sb to ovn-nbctl and hope it can persist more data before leader change to amend 1

2. Extend gc interval to amend 2

3. When it happens again log it

add filter for internal vpc

Some user upgrade form 1.6 to 1.7 by just change the image. The crds and rbacs are not updated, however the controller starts successfully and hide the issues. This check will panic kube-ovn-controller if crds are not ready and expose the issues.

auto external vpc proc