Wrong datapoints on PT2 dashboard for retried workflows

[huydhn]

When a nightly perf workflow is retried and succeeded, i.e. https://github.com/pytorch/pytorch/actions/runs/12756732787. The timestamp of the retry is newer than the first run like the case of commit 1dab79470dbecef79ba4c7d4308d8a181091e58e in https://hud.pytorch.org/benchmark/compilers?dashboard=torchinductor&startTime=Wed%2C%2008%20Jan%202025%2018%3A10%3A31%20GMT&stopTime=Wed%2C%2015%20Jan%202025%2018%3A10%3A31%20GMT&granularity=hour&mode=inference&dtype=bfloat16&deviceName=cuda%20(a100)&lBranch=main&lCommit=1dab79470dbecef79ba4c7d4308d8a181091e58e&rBranch=main&rCommit=f295eff5121bb4faa5abcc9c2bac0f61ae68cc3c. We should always the timestamp of the first run.

cc @yangw-dev

[yangw-dev]

@huydhn you mean always show the timestamp of the first run? what is the expected first run info