Liveliness checker has a false positive result in case of high load

Description

The problem only occurs for UDP traffic.

Steps to reproduce:

1. Create kind cluster
2. Deploy Kernel2Ethernet2Kernel
3. Install iperf3 on the NSC/NSE
4. Start iper3 on the NSE: iperf3 -s
5. Start iper3 on the NSC: iperf3 -c 172.16.1.100 -u -b 10G
6. Output:

...
[  5]   5.00-6.00   sec   878 MBytes  7.37 Gbits/sec  660633  
[  5]   6.00-7.00   sec   880 MBytes  7.38 Gbits/sec  662190  
iperf3: error - unable to write to stream socket: Network unreachable

7. NSC log:

Nov 10 08:16:16.942 [ERRO] [id:alpine-2a483148-ea71-4cc0-a3ac-22bb1e8e9a75-0] [dstIP:172.16.1.100] [srcIP:172.16.1.101] [type:networkService] (7.1)         No packets received
Nov 10 08:16:16.942 [WARN] [id:alpine-2a483148-ea71-4cc0-a3ac-22bb1e8e9a75-0] [heal:eventLoop] [type:networkService] (7.2)         Data plane is down
Nov 10 08:16:16.942 [DEBU] [id:alpine-2a483148-ea71-4cc0-a3ac-22bb1e8e9a75-0] [heal:eventLoop] [type:networkService] (7.3)         Reconnect with reselect

Start by reproducing the UDP case in a kind cluster with Kernel2Ethernet2Kernel and iperf3 at 10G, then correlate the NSC liveliness-checker logs with the reported socket error. The work is complete when sustained UDP traffic no longer produces a false “No packets received” result or an unnecessary data-plane reconnect.