Better max autotune discovery on pt2 dashboard

[huydhn]

Max autotune benchmark run only once per week due to its cost, so it only shows up for some commits, i.e. https://hud.pytorch.org/benchmark/compilers?dashboard=torchinductor&startTime=Wed%2C%2001%20Jan%202025%2021%3A01%3A09%20GMT&stopTime=Wed%2C%2015%20Jan%202025%2021%3A01%3A09%20GMT&granularity=hour&mode=inference&dtype=bfloat16&deviceName=cuda%20(a100)&lBranch=main&lCommit=67f85ccdcf56894d653b4d37cd7651eefa0ddf8c&rBranch=main&rCommit=00df63f09f07546bacec734f37132edc58ccf574

It's pretty confusing because it's not clear which commit has the data.

cc @yangw-dev @zou3519