Firefox fixes

[praschke]

• the readme and CSP wasn't changed when Kibana's handlebars was added to the repo.
• FF 102 moved script and style injection to the scripting API
• instead of the Chrome-only StyleOrigin enum, string literals may be used
• window.getSelection() can return null on FF, throwing some noisy errors
• there's something wrong with declarativeNetRequest on FF that breaks after the browser restarts.

Fixes #173. Improves #96. The remaining lint warnings on FF pertain to storage.session and eval in Handlebars and WanaKana.

[djahandarie]

@praschke Very nice! Thank you. This fixes a ton of small problems and gets rid of MV2 / fallback cruft.

One question: what is declarativeNetRequest used for in the first place? I imagine it was in there for a reason, so I'm trying to figure out what we would need to track. Maybe we could open a separate issue to track fixing whatever functionality is enabled by that permission on FF?

[github-actions]

✔️ No visual differences introduced by this PR.

View Playwright Report (note: open the "playwright-report" artifact)

[praschke]

One question: what is declarativeNetRequest used for in the first place? I imagine it was in there for a reason, so I'm trying to figure out what we would need to track. Maybe we could open a separate issue to track fixing whatever functionality is enabled by that permission on FF?

declarativeNetRequest is what Chrome is forcing extensions to use with MV3. Previously webRequest was available in Chrome. Firefox is maintaining the availability of webRequest because if uBlock Origin got gimped they'd probably lose the last of their users.

declarativeNetRequest and webRequest are used in request-builder.js to make audio requests. The existing logic with webRequest is still present, but if declarativeNetRequest is detected then it uses that instead. No functionality is lost.

[djahandarie]

Got it, makes sense, thank you for elaborating! In that case LGTM!