Update arp_update to refresh neighbor entries from APP_DB

[prsunny]

- What I did
arp_update to refresh neighbor entries from APP_DB

- How I did it
Modified arp_update script

- How to verify it
syslogs

- Description for the changelog

- A picture of a cute animal (not mandatory but encouraged)

[lguohan]

retest vsimage please

[prsunny]

retest vs please

[prsunny]

retest vsimage please

[prsunny]

retest vs please

[zhenggen-xu]

In general, if we are facing some scale issue already (e,g, > 1K neighbors), we should change this arp_update scripts to scapy based (like the one here: https://github.com/Azure/sonic-swss/blob/11fe6b520693c31c5e1398274035e103d2977a31/neighsyncd/restore_neighbors.py), it is way too slow here. like arping could take ~0.1 second. ping had 0.2 second timeout. we can easily hit 100s seconds to probe the devices, and totally time could be way more than 300 seconds. In the script should should check the time clock and the sleep should exclude the time spent on the processing. (if already more than 300s, no sleep).

[zhenggen-xu]

Approving this PR with some comments:
1, We should continue to root cause of the netlink lost issue. The changes in this PR could be used to monitor such cases so we can cross check events.
2, Scalability of this arp_update script is concerning, we should address that by raw socket mechanism etc, also make the update interval more accurate regardless the time of the probe process. (check time clock instead of sleep blindly).