Consider more reliable and stable detection methods for distinguishing zimit classic and zimit2

[Jaifroid]

As suggested here, we could look for 'warc2zim' AND 'zimit' strings in Scraper metadata (we currently only look for 'warc2zim', but it's not currently guaranteed to be stable), and if '_sw:yes' is not in tags, then it's zimit2. If it's there, then there is a Service Worker, meaning it's zimit classic.

We currently rely on finding 'warc-headers' in the declared MIME type. But it's possible (if currently unlikely) that such headers could be reintroduced if they are needed in future versions of zimit2, so it would be good to have other options as outlined above.