[PoC] Live Text image analysis on macOS

[rcombs]

This is a draft for discussion; this feature is not mergeable in its current state.

This adds support for the Live Text API on macOS, which allows the user to select text within a video.

In this demo, I select some Japanese text from a video file:

resized.mov

The copied text is:

トゲナシトゲアリ「雑踏、僕らの街」
Produced by 玉井健二
作詞・作曲：大濱健悟
編曲：玉井 健二，大濱 健悟
Product of agehasprings

Which is largely correct modulo spacing.

This will need a number of changes before it's mergeable, some of which I'd like to get some discussion started on:

• The API calls required to capture the image and convert it to a form usable by the system APIs are largely hacked in haphazardly right now; I'm not sure what the best solution for some of this is
• Currently, I'm capturing a window screenshot (so OSD is included), but I'm not informed when the OSD updates, so it becomes outdated easily; the simplest solution might be to simply not support the OSD (which would mean taking a subtitles screenshot and configuring the overlay view to be aware of the video's margins within the window)
• This will presumably want to be gated behind a setting
• Long-term, [WIP/POC] Add API to obtain metrics and shape data libass/libass#856 should provide the text metrics we'd need to implement our own selection functionality for text drawn using libass, at which point we'd want to switch this to use a video screenshot
• This reuses some image conversion utility routines out of screenshot: add screenshot-to-clipboard command #15568, pulled out into their own new file; that'll need to be reconciled once either feature lands
• The system only analyzes text in the user's configured languages by default; we should grab the list of languages that could plausibly be in the video or displayed subtitles (video stream language, all audio stream languages, and selected subtitle stream language seems like a reasonable set?) and signal those to the analyzer

[Akemi]

This will presumably want to be gated behind a setting

yeah an option that can be toggled at runtime, so it doesn't interfere with window dragging if not wanted.

[edit]
the user could configure the behaviour that way and we don't need to hardcode anything (like on pause). eg auto-profile on pause to set this option.

[Akemi]

since ImageAnalysisOverlayView is added as a subview all the ImageAnalysisOverlayView functionality/delegate/as much as possible should be moved into its own view class if possible.