[benchmark] Improve printed results

[oliviertassinari]

Use https://necolas.github.io/react-native-web/benchmarks/ as inspiration:

• Display the standard deviation
• Increase the number of runs from 10 to 15
• Create a test case array for simplicity
• Turn into % display to be more CPU agnostic (help with collaboration)

noop (baseline):
  04.59 ±00.10ms
React primitives:
  44.05 ±02.19ms
React components:
  126 ±8%
Styled Material-UI:
  174 ±6%
Styled emotion:
  163 ±6%
Styled SC:
  191 ±5%
makeStyles:
  168 ±6%
Box Baseline:
  180 ±4%
Box Material-UI:
  377 ±6%
Box Theme-UI:
  316 ±6%
Box Chakra-UI:
  269 ±9%
styled-components Box + @material-ui/system:
  358 ±8%
styled-components Box + styled-system:
  307 ±9%

@mnajdova Could you run the script locally and commit the printed output in the markdown so we can evaluate if it's stable or not (on this PR)?

[mui-pr-bot]

No bundle size changes

Generated by 🚫 dangerJS against 9671c4c

[mnajdova]

@mnajdova Could you run the script locally and commit the printed output in the markdown so we can evaluate if it's stable or not (on this PR)?

@oliviertassinari here it is 980a9c8

[oliviertassinari]

Thanks, it's not very stable, it would benefit from more. However, we have enough precision to clearly see the difference of performance between the different options. Maybe if it was running in the CI it would have a smaller std.

[mnajdova]

The info is def more useful than previosly 👌