[Migration] Update dashboard to process results from openxla-benchmark

[pzread]

Benchmark results uploaded from openxla-benchmark are a little different from iree-samples.

Here is the example from T5_LARGE_FP32_JAX benchmarks:

From iree-samples: https://gist.github.com/pzread/2be3b2db7c0ffa14518085f08e33b815
From openxla-benchmarks: https://gist.github.com/pzread/cf3e089d0a4e0d5c452aebc1df6dbc44

The major differences are:

• Changes in benchmark id and benchmark name
• Changes in some metadata values (e.g. data_type and device field)
• Changes in the format of python_environment field

I think the pain point will be changing the benchmark id, which means we might need a backfill of the historical data. We can still show new data in the dashboard first then backfill the previous data

[pzread]

@beckerhe temporarily assigned this to you. Let me know if you have time to work on this : )

[pzread]

So the format of the benchmark id needs to be finalized first (so we won't change it later...). This is blocked by #41

[pzread]

As an offline discussion, I'll create a new GCS bucket to store new benchmark results #20 (currently the new benchmark results are uploaded to postsubmit GCS bucket with random job id, which is not well-organized). Then we can start testing and doing data migration with it

[pzread]

All benchmarks are running and uploading results to GCS, so I think we can start importing them into the dashboard.

Here is an example of the benchmark results gs://comparative-benchmark-artifacts/2023-07-12.1689195682

[beckerhe]

Awesome. I will prepare the switch.