Re-enable ICMP responder and gratuitous ARP service for active-standby dualtor topologies.

[vivekverma-arista]

Description of PR

Summary:
Fixes #119

Type of change

• Bug fix
• Testbed and Framework(new/improvement)
• Test case(new/improvement)

Back port request

• 201911
• 202012
• 202205
• 202305
• 202311

Approach

What is the motivation for this PR?

#221 introduced oscillation logic in active-standby dualtor. These oscillations happen continuously in the testing environment as well because we don't run ICMP responder as it was disabled some time back in #9117.

These continuous oscillations interfere with the testing and has made a lot of traffic tests flaky in active-standby dualtor.

Details can be found in #119

How did you do it?

Re-enabled ICMP responder and gratuitous ARP service in active-standby dualtor topologies.

We have introduced new fixtures toggle_all_simulator_ports_to_rand_selected_tor_unconditionally and toggle_all_simulator_ports_to_enum_rand_one_per_hwsku_frontend_host_unconditionally similar to active-active dualtor to run few tests in active-standby mode where ICMP responder interferes with the testing and we need to pause it.

This also affects then following tests which have been fixed

1. everflow tests: These tests shutdown BGP on the randomly selected ToR and if ICMP responder is running toggle_all_simulator_ports_to_rand_selected_tor fails to toggle the MUX direction towards this ToR because it is now technically unhealthy as it lost routes ( which should be the expected behaviour in the production environment as well ). The fix for this test is very similar to active-active dualtor by using the fixture toggle_all_simulator_ports_to_rand_selected_tor_unconditionally. If ICMP responder is not running then toggle_all_simulator_ports_to_rand_selected_tor successfully toggles the MUX direction towards the unhealthy ToR which is a testing gap that got introduced due to disabling of ICMP responder.

2. In case of pfcwd/test_pfcwd_function.py and arp/test_unknown_mac.py. ICMP responder and GARP interferes with the testing therefore the fix is to selectively pause them for these tests and run them in active-standby mode by using the new fixtures.

How did you verify/test it?

Tested on Arista-7260 and Arista-7050 platforms with dualtor and dualtor-120.

Any platform specific information?

Supported testbed topology if it's a new test case?

Documentation

[vivekverma-arista]

This has been fixed by modifying the product code therefore closing this pull request: sonic-net/sonic-linkmgrd#250