Info extractor rework

[dyc3]

The info extractor has some problems that make implementing new features hard.

• URL parsing is too tightly coupled to metadata extraction.
• There isn't a clean way to obtain the URL that was resolved into a Video or array of Video
• Collections are not as clearly defined as Video is, resulting in lots of duplicate code.
• ServiceAdapters don't have a clean, easy to follow mechanism to provide Videos where service != adapter.serviceId
• There isn't a clean, super type-checked way to provide startAt and endAt metadata that is embedded into the url (eg. youtube's t=5s query parameter for starting the video at 5 seconds)
• "highlighting" a video in an add preview is incredibly janky, (and i think its broken right now)
  • highlighting is a feature that shows the referenced video at the very top of the list if you put in a youtube link with a playlist and a video referenced (eg. https://www.youtube.com/watch?v=OtqD0ddbLic&list=PLqKpXXbHNvC83j0C8UsR817qcrXupLGZ5)

I want the new info extractor to be able to:

• resolve all the above issues
• do pagination (Add Preview Pagination #172)
• provide extra metadata for queue items (eg startAt and endAt)
• maintain all the caching and bandwidth saving logic we already have
  • maybe make it possible to cache add previews locally
• be able to cache collections (playlists, channels, etc.)

So far, what I'm thinking, when diagrammed out, is this:

classDiagram

    class VideoId {
        service: string
        id: string
    }

    class Video {
        id: VideoId
        meta: VideoMetadata
    }

    Video <-- VideoId

    class CollectionId {
        service: string
        id: string
        page: string
    }

    class ExtractStrategy {
        <<abstract>>
        resolve(url) Video[]
    }

    ExtractStrategy <|-- Single
    ExtractStrategy <|-- Multi
    Multi <|-- Collection

    class ServiceAdapter {
        <<abstract>>
        canHandleUrl(url) boolean
        isCollectionUrl(url) boolean
        parseSingle(url) VideoId
        fetchSingle(videoId) Video
        fetchMulti(videoId[]) Video[]
        parseCollection(url) CollectionId
        fetchCollection(collectionId) VideoId[]
    }

    Youtube <|-- ServiceAdapter
    Vimeo <|-- ServiceAdapter
    Direct <|-- ServiceAdapter

Loading

This isn't perfect or complete, and I'll be refining it and updating this issue.

Some notes:

• In CollectionId, page is a string so that it can either be a number or a page token (youtube api uses those)
• ServiceAdapter should provide a default, overrideable impl for fetchMulti(), allowing derived adapters to override it with something more optimized if applicable.