Container probes fail because of NetworkPolicy dropping them

[tnqn]

Describe the bug
If a Pod has readinessProbe defined, and there's Ingress NetworkPolicy selecting the pod but the allowed rule doesn't include the readinessProbe port, the Pod would be marked as NotReady forever because the readinessProbe traffic are all dropped.

This is caught by K8s test case

[Fail] [sig-network] NetworkPolicy [LinuxOnly] NetworkPolicy between server and client [It] should enforce updated policy [Feature:NetworkPolicy]

To Reproduce

1. Define a Pod with two containers which listen on 80 and 81 separately, enable ReadinessProbe on 80 and 81 correspondingly.
2. Create a service for the above Pod.
3. Create a NetworkPolicy which allows clients to access port 80 and applies to the above Pod.
4. Clients can't access the Pod via the service because the Pod's readiness probe fails and it is marked as NotReady.

Expected
The readiness probe should success regardless of what policy is defined.

Actual behavior
The readiness probe failed if the ingress rule doesn't include all ports being probed.

Versions:
Please provide the following information:

• Antrea version (Docker image tag).
  0.1.1

Additional context
PR #104 appends the NodeIP to FromAddress of all NetworkPolicy rules, but it doesn't handle the case that the rule itself doesn't cover all probed ports.
@jianjuns proposed to add a flow to make traffic from Node bypass NetworkPolicy flows, which should be the proper approach.

[tnqn]

Will revert #104 after #175 is merged