We don't require npm to be installed

The old version is no longer maintained and doesn't support rollup v3

In greenlight we copied this code for scanning images, either from
real image files, or from PDF files converted to images via pdf.js.

To avoid duplication clean up the internal API and expose it.

Also write some unit tests for the scan engine while at it.

Multiple changes:

The worker can now be auto-bundled via rollup and no longer has to be passed manually.
This is handled via a js script file which builds a blob containing bundled code.
This requires some CSP exceptions in the target applications, so mention that in the README.

scanImage() now takes an option object and defaults to returning more information, so
handle both changes (returnDetailedScanResult now defaults to true)

Component users can now drop their worker copying code and need to adjust their CSP.

1.4 and newer no longer detect any QR codes for some reason.
Needs to be investigated why..

It has new API which we want to use.

Newest Ubuntu switched Firefox to a snap, which breaks karma.

Try to use playwright to download and cache firefox/chromium
and use those for karma.

Compared to puppeteer where we did this some time ago this caches
the browsers globally, and only downloads them when we start the tests.

This reverts commit 6c632dc6