Investigate and re-enable the OTA CI test

[selissia]

Problem

OTA CI tests were disabled in #15424 because of instability, possibly related to mDNS. Slack discussion: https://csamembers.slack.com/archives/G014G30SVV0/p1645552481568699

Need to investigate what's happening and eventually re-enable the CI.

Proposed Solution

[mykrupp]

When Linux fails, the requestor does not commission.

When Darwin fails, send ota announce provider command fails

Sometimes github also fails to clone some repos

[mykrupp]

#15659

[mykrupp]

The OTA test on linux platform fails about 20% of the time.
When the test fails, it is because the requestor does not commission.
The requestor doesn't commission because chip-tool erroneously finds the port corresponding to the provider instead of the requestor in these cases.

Some logs when commissioning succeeds
CHIP:TOO: Discovered Device: fe80::4402:baff:fee7:5200:5540 when commisioning provider
CHIP:TOO: Discovered Device: fe80::4402:baff:fee7:5200:5560 when commissioning requestor

Some logs when commissioning fails
CHIP:TOO: Discovered Device: fe80::4402:baff:fee7:5200:5540 when commissioning provider
CHIP:TOO: Discovered Device: fe80::4402:baff:fee7:5200:5540 when commissioning requestor

The requestor is given the port corresponding to the provider, causing requestor commissioning to fail.

It needs to be determined why MDNS sometimes resolves the incorrect port for the requestor.

[mykrupp]

The incorrect port is picked up here -

connectedhomeip/examples/chip-tool/commands/pairing/PairingCommand.cpp

Line 203 in 3ee4f4d

ChipLogProgress(chipTool, "Discovered Device: %s:%u", buf, port);

[mykrupp]

Chip-tool and requestor logs upon failure:
chip-tool-commission-requestor.txt
requestor-log.txt

[selissia]

The Requestor UDP port is always set to 5560 by using the the command line option --secured-device-port 5560 .

[mykrupp]

I tested a few solutions suggested by other Matter developers 

1. Generate a random UDP_PORT and DISCRIMINATOR to make the code hermetic.
2. put rm -r /tmp/chip_* back in the code.

This did not fix the unstableness of the test. 

[bzbarsky-apple]

@carol-apple Is this still an issue given the YAML coverage you added?

[carol-apple]

#18344 fixes this issue