A few BookFinder improvements (including a fix for #2238)

[mikiher]

The following improvements were implemented:

• Handle "abridged/unabridged" in titleCandidate generation
• Clean up "et al[.]" in author normalization.
• Sort Audible match results by ascending duration difference
  • this required an API change, since up until now /search/books had no context (it only accepted title, author, and provider). We now pass the libraryItem id from the client, retrieve it from the database, and pass it to BookFinder.search().
  • This change is fully backwards-compatible.
  • This change is also generally a good preparation for future context-based matching improvements (e.g. prefer results with same narrator)
  • I believe this resolves Sort Book Match Results by closest to length #2238

[advplyr]

Works great, thanks!

[advplyr]

#2517 there is a case I didn't think of where audible returns a bunch of results that are bad matches but if one of those matches has a closer duration it will show up at the top.
I'm not sure the best resolution for this. Maybe we have a max levenshtein distance or something.

[mikiher]

So, there's a filterSearchResults function that tries to do this (but only for OpenLib results).
I tried to enable it for other providers at some point, but it had some issues and I ditched the idea at the time (I don't remember the exact details).

I'll look into it once more, and perhaps change the current implementation so it works well for all providers.

[mikiher]

But, just to be clear, even if it is better implemented, I think filterSearchResults is just a patch.

I think that for the purpose of running Quick Match on a large scale, what we really need is to have a confidence score for the top result. If it is above some threshold, we automatically match, if not, we mark it and let the user review - I know I want this feature badly... it's the only way I'm going to be able to get some order into my library :)