Re-plot performance testing based on individual files

[asteiker]

The plot shown in #19 demonstrates collated testing runs across i/o libraries for each data format (original, repack, kerchunk-original, kerchunk-repack). It would be more valuable to create a scatter plot of the performance testing for individual file runs based on #20 to better observe the within-group variability. Grouping by tool may be most valuable.

Tasks

Beta Give feedback

1. Update h5cloud/helpers/s3filelinks.json to point to persistent data with new bucket
2. Re-run /h5cloud/notebooks/run-tests.ipynb with a new results directory
3. Persist pandas dataframe in benchmark notebook to write to csv
4. Replot as scatter plot grouped by tool
5. Plot all file and i/o param combinations on a single plot

[asteiker]

Dependent on #28