Estimate the impact of using the plugin

[maudnals]

Compare the Lighthouse audit for a realistic app with vs. without this plugin.
See if there are any unexpected consequences of using the plugin.

[maudnals]

Done. Audited the forked application react-real-world-example-app.

Full results:

• Audit without the strict CSP plugin
• Audit with the strict CSP plugin

Results summary:

• Without the strict CSP plugin, perf audit = 22
• With the strict CSP plugin, perf audit = 54 (🟢 better). Because all scripts became async, since we added them dynamically.
• (all other audits are the same)

Conclusion:

• This plugin does no performance harm in this case.
• This little experiment doesn't cover the case with defered scripts. As of this version, using the plugin would have an unexpected impact if the application had defered scripts. We need to take Feature Request: Automatically defer sourced scripts in the script loader #12 into account.

[maudnals]

One way this could be an issue:
sourced script
then inline that uses some stuff from the sourced script

So, with our plugin:
the inline script will be left
but the sourced script will be loaded async
so inline will load before the sourced one <-- this may be an issue.

For SPAs, where everything's on the bundle.
GA and others are not dependent on order of each other

can be fixed with manual rewriting