DrCI fails to identify test as broken on trunk

[malfet]

From pytorch/pytorch#104214 DrCI identifies test_sparse_semi_structured as NEW failure, even though it is caused by trunk failures introduced by pytorch/pytorch@cf5262a that was later reverted

Hud link: https://hud.pytorch.org/pytorch/pytorch/pull/104214?sha=dd52521aa51e4d609fb97b6a4cf657fc89c49308

[huydhn]

Just a note, this discrepancy happens because the failed test was run in 2 different shards, thus simply comparing job names between didn't match

[huydhn]

There is another issue with comparing job names is that the suffix like unstable is also included. As the suffix is added dynamically by checking the open/close status of the corresponding unstable issue. This also leads to discrepancy, i.e. pytorch/pytorch#104067 has a new failure for linux-bionic-cuda12.1-py3.10-gcc9 / test (deploy, 1, 1, linux.4xlarge.nvidia.gpu) (gh). Its base commit as the same failure https://hud.pytorch.org/pytorch/pytorch/commit/67babf7a4514f38a1f96fbaf947d357097c2c4d8 but has the suffix linux-bionic-cuda12.1-py3.10-gcc9 / test (deploy, 1, 1, linux.4xlarge.nvidia.gpu, unstable) because pytorch/pytorch#103654 was open at that time