Antrea Traceflow should validate that destination port isn't empty when destination is a K8s Service

[Atish-iaf]

Describe the bug
Antrea Traceflow should validate that destination port isn't empty when provided destination is a K8s service.
If destination port is provided, it works fine but if it is missed it just uses destinaton Service IP(clusterIP) and runs Traceflow as Pod-to-external IP ICMP case which is incorrect.

To Reproduce
Run Antrea Traceflow with source as a Pod and destination as a K8s Service
antctl tf -S ns1/deployment1-cb9cdd4bd-62dbp -D ns1/svc2

Expected
It should return error that destination port is required when destination is a service.

Actual behavior

name: ns1-deployment1-cb9cdd4bd-62dbp-to-ns1-svc2-jtkgwnz4
phase: Succeeded
source: ns1/deployment1-cb9cdd4bd-62dbp
destination: ns1/svc2
results:
- node: antrea-2
  timestamp: 1722949239
  observations:
  - component: SpoofGuard
    action: Forwarded
    srcPodIP: 192.168.2.5
  - component: Forwarding
    componentInfo: Output
    action: ForwardedOutOfOverlay

Additional context
Also, antctl tf should validate that flow arg shouldn't be empty when destination is a K8s service.

[antoninbas]

Should we use a reasonable default instead? For example, we could use the first protocol / port in the Service's ports list. Default would be populated in the Traceflow controller in the Antrea Agent.

On the antctl side, we could issue a warning but still proceed with the Traceflow request.