Add PDF worker support

[mvasilak]

• Adds support for PDF worker via PDFWorkerWebViewHandler.
• Adds PDFWorkerController that exposes PDF worker for extracting recognizer data and full text.
• Adds RecognizerController that extracts recognizer data, performs remote recognizer request, and finally performs identifier lookup to produce an item response, for selected local PDF file. (Logic for remote recognizer response processing is coped from https://github.com/zotero/zotero/blob/7b7a0b9e4692a0ca32984b763f67b7b10edf96fd/chrome/content/zotero/xpcom/recognizeDocument.js#L411)
• Uses RecognizerController in share extension, to create parent item for a shared local PDF file.

TODO:

• Use recognizer with Retrieve Metadata action for standalone attachments.
• Use recognizer for standalone files added within the app.
• Show in UI that there is metadata retrieval in progress for affected items.

To be considered:

• Add popup window with all metadata retrievals' progress, that will be minimized in the items bottom toolbar.
• Allow canceling metadata retrievals.

[mvasilak]

@michalrentka @dstillman this is ready for a first review / user testing. I did a very simple UI iteration, where items that are retrieving metadata, are indicating this by an animating subtitle, as it is empty for attachment type. If functionality is as desired, we can improve the UI as we see fit.

[michalrentka]

Didn't go through everything yet, but there will probably be bigger changes, so I'll have to go through most of it again anyway.

[michalrentka]

I think this is good to submit now. Although ideally I'd like to see some changes in RecognizerController, but they would require some further refactoring which might be unnecessary time kill at this point.

Anyway if I understand RecognizerController correctly, the order of events there is one big chain, which could be summed up as: enqueue pdf worker -> start remote recognition -> lookup identifiers -> process identified items (create new items).

So theoretically it could be simply written as something like:

pdfWorkerController.queue(work: PDFWorkerController.PDFWork(file: task.file, kind: .recognizer))
.filter({
    // filter out in-progress states
})
.flatMap({
    // Report errors if failed/cancelled, continue with extracted data
})
.flatMap({
    return apiClient.send(request: RecognizerRequest(parameters: data))
})
.flatMap({
    // extract identifiers
})
.flatMap({
    // This would require a new controller which would loop through all identifiers and report when they are all finished
    // It could be `Observable` and report multiple `onNext` events so that each event is handled separately if needed, but I guess it would be even better to gather all data first and then have one database write with multiple items
    identifierController.lookupIdentifiers()
})
.subscribe(
    // Process lookup results - create/report items
    // Report errors
)
.disposed(by: disposeBag)

It would be neater and much easier to understand quickly what's going on. Of course there would probably have to be some more filters/flatMaps for special cases, but I think it would still be even less verbose than it is currently, where you have to subscribe to and process each individual step.

Anyway this is not required at this point, just a thought for the future.