Continue when a workload is not working

[rmahdav]

This PR fixes issue #82.

If a workload throws an error (during initialization or running) or timeout, we show a warning on the result page and notify the user. I have tested the following situations:

1. A number of workloads throw error in each iteration. The list of them are shown on the result page and because they were not completed in any of the iterations, we exclude the measurement (in some cases is a few nano seconds because of the possible successful steps) from the result list and geomean calculation:

2. If a workload fails in some iterations, we show it as part of the failed workloads but include the successful iterations in results list and geomean calculation. The feature extraction on webgpu failed on two of the iterations in the below screenshot:

We can see the number of the iterations in the shown table for the workload:

I also tested if the heights of the warning box is dynamic based on the number of the failed workloads and if it is not shown if all the workloads are successful.

This change has two possible issues which we can fix in follow-up PRs, if we decided to fix them:

• When some of the workloads are skipped, the current progress bar for each iteration does not perform accurate. You can see that the last iteration is completed in the UI, but workloads are still running. IMO, it is small enough issue to ignore for now.
• The warning box moves with the scrolling. It may be better if it is fixed.

[rmahdav]

This PR does have conflict with PR #114 , I'll wait for that one to merge and then rebase this one on top of that.