OVN didn't follow upstream ovs code and use submodule to preserve a stable ovs codebase.
If build with upstream ovs, some functions has different args and the build failed. So we
also need to follow the ovn build way.

s thing!

use UpdateStatus method instead of patch

1. add e2e testing for 3-master cluster;
2. add e2e testing for StatefulSet Pod force deletion.

When node is not ready lots of pods may stay in `Terminating` status. The inspection previously will always put the pods into addQueue which will waste lots of cpu.

1. remove kubectl
2. merge all go binary into one